You are on page 1of 422

GROUNDWATER: MODELLING,

MANAGEMENT AND CONTAMINATION

No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
GROUNDWATER: MODELLING,
MANAGEMENT AND CONTAMINATION

LUKA F. KÖNIG
AND
JONAS L. WEISS
EDITORS

Nova Science Publishers, Inc.


New York
Copyright © 2009 by Nova Science Publishers, Inc.

All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.

For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com

NOTICE TO THE READER


The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.

Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.

This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA


Groundwater : modelling, management, and contamination / Luka F. König and Jonas L. Weiss
(editors).
p. cm.
ISBN 978-1-60741-395-0 (E-Book)
1. Groundwater flow--Mathematical models. 2. Groundwater--Pollution. I. König, Luka F. II.
Weiss, Jonas L.
GB1001.72.M35G764 2008
551.4901'5118--dc22 2008025050

Published by Nova Science Publishers, Inc. New York


CONTENTS

Preface vii
Short Communication 1
Use of Artificial Neural Networks 3
for Predicting of Groundwater Contamination
Goloka Behari Sahoo and Chittaranjan Ray
Research and Review Studies 15
Chapter 1 GIS-Based Aquifer Modeling and Planning Using Integrated 17
Geoenvironmental and Chemical Approaches
Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios
Chapter 2 A Modeling Approach Supporting Air Sparging System Design 79
Sabrina Saponaro, Sara Puricelli, Elena Sezenna
and Luca Bonomo
Chapter 3 Groundwater Interactions with Surface Waters: Consequences on 113
Diffuse Pollution Pathways
A. Azzellino, R. Salvetti and R. Vismara
Chapter 4 Modelling of Water-Solid Interactions: 133
A Discussion of Different Approaches
Marek Šváb and Lenka Wimmerová
Chapter 5 Fundamentals of Groundwater Modelling 149
Husam Baalousha
Chapter 6 Contaminants in Groundwater 167
and the Subsurface
Z. Yu, Y. Huang, A. Baron, X. Chen, C. Yang,
D. Kreamer and M. Johnson
Chapter 7 Preliminary Studies for Designing a Wetland for Arsenic 187
Treatment
Raja Chowdhury, Defne Apul and Daryl Dwyer
vi Contents

Chapter 8 Groundwater Management in the Northern Adriatic Coast 203


(Ravenna, Italy): New Strategies to Protect the Coastal Aquifer
from Saltwater Intrusion
Beatrice M.S. Giambastiani, Pauline N. Mollema
and Marco Antonellini
Chapter 9 Hydrogeology and Geochemistry 231
of the Friuli Venezia Giulia Plain Alluvial Aquifers, Northeastern
Italy
Franco Cucchi, Giuliana Franceschini and Luca Zini
Chapter 10 Analytical and Numerical Solutions for Convection-Diffusion- 259
Dispersion-Reaction Equations with General Initial-Conditions:
Theory and Applications
Jürgen Geiser
Chapter 11 Flow Simulation and Optimal Management 291
of Groundwater Resources. The Balance
between Accuracy and Computational Efficiency
K.L. Katsifarakis
Chapter 12 Ground-Penetrating Radar in Ground 309
Water Studies
Nigel J. Cassidy
Chapter 13 A Note on the Propagation of Water Table Waves: 351
Dual Length Scale Considerations
Nick Cartwright, Peter Nielsen, David P. Callaghan
and Ling Li
Chapter 14 Distribution and Source of Nitrogen Compounds in the 361
Groundwater of Kuwait
A. Akber, A. Mukhopadhyay, E. Azrag, E. Al-Awadi,
A. Al-Haddad and H. Al-Qallaf
Index 393
PREFACE

Groundwater (GW) is one of the most valuable natural resources and for that reason, the
GW protection and management is vital for human evolution, socio-economic development
and ecological diversity. During the last decades, the continuously increasing need of water
has led to a rapidly growing awareness in the field of GW management. At the same time
over exploitation and pollution of water resources are threatening the ecosystems. The
combination of these two problems which have acquired worldwide dimensions has forced
many scientists working in relative fields to search new, multidisciplinary approaches to
address them. Effective management and protection of groundwater resources require detail
knowledge and quantitative/qualitative characterization of aquifers. Thus, modeling and
planning of the GW through the use of modern technologies and approaches have become of
high priority towards this direction. This book provides leading-edge research on this field.
Artificial neural networks are empirical mathematical tools proven to represent complex
relationships of hydrological systems. Neural networks are increasingly being applied in
subsurface modeling where intricate physical processes and lack of detailed field data prevail.
Two types of ANN models: Back propagation neural network (BPNN) and radial basis
function neural network (RBFN) are examined to predict the pesticide contamination of
domestic wells. Because sample collection, analysis, and re-sampling are expensive, a large
dataset is not available for ANN use in this study. This Short Communication presents
analyzes of raw data for preparation of input subsets for ANN use. Thus, a clustering
technique is used to divide the whole dataset into three subsets: training, validating, and
testing. The sensitivity analysis was carried out by deleting one or more input variables from
the input data set to measure the importance of one variable over the other in terms of ANN
prediction performance. It provides a sense of the effect of each parameter on pesticide
occurrence in a well. The well depth, depth to aquifer material from land surface, and on-site
pesticide storage are found to be important parameters in pesticide occurrence in well.
Groundwater (GW) is one of the most valuable natural resources and for that reason, the
GW protection and management is vital for human evolution, socio-economic development
and ecological diversity. During the last decades, the continuously increasing need of water
has led to a rapidly growing awareness in the field of GW management. At the same time
over exploitation and pollution of water resources are threatening the ecosystems. The
combination of these two problems which have acquired worldwide dimensions has forced
many scientists working in relative fields to search new, multidisciplinary approaches to
address them. Effective management and protection of groundwater resources require detail
viii Luka F. König and Jonas L. Weiss

knowledge and quantitative/qualitative characterization of aquifers. Thus, modeling and


planning of the GW through the use of modern technologies and approaches have become of
high priority towards this direction. This book provides leading-edge research on this field.
Groundwater (GW) is one of the most valuable natural resources and for that reason, the
GW protection and management is vital for human evolution, socio-economic development
and ecological diversity. During the last decades, the continuously increasing need of water
has led to a rapidly growing awareness in the field of GW management. At the same time
over exploitation and pollution of water resources are threatening the ecosystems. The
combination of these two problems which have acquired worldwide dimensions has forced
many scientists working in relative fields to search new, multidisciplinary approaches to
address them. Effective management and protection of groundwater resources require detail
knowledge and quantitative/qualitative characterization of aquifers. Thus, modeling and
planning of the GW through the use of modern technologies and approaches have become of
high priority towards this direction.
Despite advances in modeling tools and geoenvironmental and chemical methods, aquifer
characterization remains an extremely difficult problem due to spatial heterogeneity, temporal
variability and coupling between chemical, physical, and biological processes. Solution to this
problem involves multidimensional data integration. The concept of data fusion involves the
merging of multiple data types to develop more reliable predictive models and to address
basic and applied scientific questions concerning GW modeling.
With the advent of powerful computers and the advances in geoenvironmetal methods
and space technology, efficient techniques for water management have evolved. Remote
sensing (RS), geographic information system (GIS), innovative geophysical methods
(electrical, electromagnetic and seismic) and GW flow simulation codes are such techniques
of great significance. These techniques, with the aid of chemical analyses, have
fundamentally reassigned the ways to manage natural resources in general and water
resources in particular.
The combined application of geophysical methods and chemical analysis of soil and
water samples can provide detailed information about aquifer hydraulic parameters and
possible contamination of the GW with the highest possible resolution. This information fully
integrated with other data derived either from RS (such as land use, land cover, tectonic,
bedrock feature) or existing data in map / time series / tabular form (e.g. hydro-
meteorological, borehole data) in a GIS environment can be used for GW modeling after the
application of a simulation code (such as MODFLOW). After the model has run, the results
of the simulation (estimation of GW potential and prediction of aquifer response to
groundwater pumping and recharge and the flow path analysis) can be exported into a GIS for
post processing (PP). The PP step allows for data layers to be developed and displayed in map
forms, allowing easy examination and interpretation of the results in their spatial context.
Finally, for each data layer, a weight relative to its importance can be defined and a decision
support system (DSS) for groundwater management can be evolved.
The overall results in Chapter 1, demonstrate that the application of GIS in conjunction
with the aforementioned geoenvironmental technologies and chemical analysis provide a
powerful tool to study groundwater resources and design a suitable exploration and
management plan.
Air Sparging (AS) is a remediation technique for groundwater based on the injection of
pressurized air under the water table to strip the volatile organic pollutants from the saturated
Preface ix

zone; moving in soil towards the ground surface, pollutants are captured by a Soil Vapor
Extraction (SVE) system. The effectiveness of the treatment depends on pollutant volatility,
and air flow behavior in soil, that is affected by soil properties (intrinsic permeability,
heterogeneity, anisotropy, etc.), plant configuration (number, location, depth and screen
position of AS injection and SVE extraction wells) and values of the process parameters (AS
and SVE flowrates, AS injection cycles, etc.). Airflow in soil can be described by the
multiphase flow theory, which is based on a closed system of partial differential equations,
whose solution results in air pressure and air saturation in soil as a function of space and time;
the numerical finite difference code TOUGH2 (v. 2.0) can be used as a tool to solve the
system of equations. Chapter 2 shows the application of this tool to assess the effects of
different AS injection and SVE extraction flow rates, sparging mode, surface sealing, number,
location and screen position of AS or SVE wells for the remediation of a polluted site in the
northwest of Milan (Italy).
The interactions between groundwater and surface water are complex. Surface-waters
and groundwaters are, in fact, linked components of a hydrologic continuum. In general,
diffuse pollution in surface waters is difficult to quantify since it follows a multitude of
pathways and acts on different time scales. During rainfall events most of the diffuse
pollutant load follows the surface runoff pathways and, reaches the surface acquifers
however, a fraction of this load will follow the sub-surface runoff pathways and it will
possibly reach the surface acquifers after a certain time lag. The time scale of the sub-surface
runoff pathways is very different from the surface runoff time scale and rarely a subsurface
diffuse pollution event can be directly correlated to a specific rainfall event. This is the reason
why even though there are models that enable to simulate the groundwater-surface water
system (GW-SW), yet the effect of these interactions in terms of diffuse pollution pathways
and their correspondent effect on the quality of surface waters to date are largely unknown.
To upgrade the conceptual modeling of the “groundwater–surface water” system, a broader
perspective of such interactions across and between surface water bodies is needed.
Multidimensional analyses may help in understanding the effect of such interactions, as the
characterization of the hydraulic interface and its spatial variability.
To fully understand these interactions, modeling studies need to be coupled to sound and
robust monitoring of surface- and ground- water quality data. Modeling can be combined with
multivariate statistical techniques (e.g. factor analysis) to improve our capability to “detect”
the effect of the sub-surface runoff on the water quality of specific water courses. With
studies of this kind, it was proved that the sub-surface runoff can be a significant source of
diffuse pollutant even in dry weather conditions. Studying nitrate concentrations in different
lowland rivers, the authors found that up to 60% of the measured load could be apportioned
as sub-surface-runoff-derived.
Notwithstanding, to date, very few studies have attempted to overlay measurements with
the conceptual modeling of the GW-SW system. In Chapter 3, studies about the GW-SW
system will be reviewed, the key aspects and the research needs and challenges facing this
evolving field will be discussed outlining how site-to-region regionalization studies and
cross-disciplinary collaborations could help in improving the understanding of the complexity
of the GW-SW system.
Chapter 4 contains an explanation and discussion of two different approaches to the
modelling of water-solid interactions. The term ‘water-solid interaction’ means all
interactions that are important in the frame of remediation methods, groundwater treatment
x Luka F. König and Jonas L. Weiss

and pollution migration, including adsorption, leaching of contaminants, minerals


weathering, etc.
The discussed approaches involve an exact modelling of interactions based on chemical
reactions as well as what is known as the ‘semi-empirical’ approach, which is applicable for
very complex systems. On one hand, the exact approach is very promising for a detailed
understanding of particular processes and can be specified by a possible transfer of results
among various systems. On the other hand, the semi-empirical approach can also be a simple
alternative for modelling very complex systems.
The calculation of the equilibrium solubility of zinc and copper in water under various
conditions (with and without ammonia as a complexation agent), including detailed speciation
of both metals complexes, is an example of the exact modelling. The influence of ammonia
nitrogen on the solubility of both metals is obvious: a local solubility maximum occurs at the
pH of 9.4. The results have been confirmed by equilibrium batch experiments with zinc and
copper hydroxides that have proven the local solubility maximum at the pH of 9.4.
The exact modelling is further demonstrated by the adsorption of an organic contaminant
from water on three adsorbents in a mixture demonstrating the distribution of the contaminant
between various adsorbents. Although these isotherms are not exact descriptions of particular
sorption reactions, the method of calculation demonstrates the principles of use of the exact
geochemical modelling for calculations of competitive sorption on different adsorbents.
As an example of the semi-empirical approach, a mathematical model aimed at predicting
the course of a continuous soil flushing process by use of the input data obtained from simple
batch laboratory experiments is described. An objective of the example is to apply this new
model to soil polluted by zinc and copper (11 949 mg/Kg-1 and 1 895 mg/Kg-1, respectively)
by flushing this soil with an ammonia nitrogen solution with the pH of 9.4. The model
predicts correctly the period of time needed for the removal of weakly-bound metal fractions,
as well as estimates the overall removal efficiency of metals from the soil during the flushing
process.
Using the above-described examples, the main aspects of both approaches are
demonstrated. Finally, future research possibilities are discussed.
Groundwater Modelling is an efficient tool for groundwater management and
remediation. Models are a simplification of reality to investigate certain phenomena or to
predict future behaviour. The challenge is to simplify reality in a way that does not adversely
influence the accuracy and ability of the model output to meet the intended objectives.
Despite their efficiency, models can be complicated and produce wrong results if they are
not properly designed and interpreted. Regardless of the type of model being used, similar
sequences should be followed in modelling. To help selecting the proper model, modelling
objectives should be clear and well identified.
If the conceptual model is not properly designed, all modelling processes will be a waste
of time and effort. To build a proper conceptual model, hydrogeological data should be
sufficient and reliable. Calibration and verification are the last steps in modelling before
writing the final model report.
Chapter 5 discusses the stepwise methodology of groundwater modelling with
explanation of each step. It contains a brief description of different types of models and
different types of solutions. In addition, special difficulties and common mistakes in
modelling have been discussed.
Preface xi

Remediation of groundwater contamination is becoming increasingly important as


groundwater resources are more extensively relied upon for industrial and drinking water
supplies. Subsurface contaminants can be divided into three general categories: 1) microbial
pathogens, which can be filtered out by some subsurface media and therefore pose the
greatest risk in aquifers with large cracks or conduits, 2) inorganic compounds such as major
anions and cations, metals, and radioactive compounds, whose mobility in groundwater
depends on their interaction with the geologic material, and 3) organic chemicals, including
many pesticides and nonaqueous phase liquids such as petroleum products and solvents. The
movement of contaminants through the subsurface is governed by hydrological, physical,
chemical, biological and transport processes, all of which are explained in detail.
Investigation of groundwater contamination should begin with noninvasive techniques,
including review of historical records, maps and photographs, and some geophysical methods.
Contaminant plumes can then be delineated using soil samples and monitoring wells.
Numerical models, as long as they are developed via a rigorous procedure based on specific
questions and available data, can integrate spatial and temporal variables to more accurately
describe the complexity of subsurface contaminant fate and transport. Various unsaturated
and saturated zone models are described. Remediation in its most basic forms consists of
source containment, excavation of soil in the unsaturated zone, and pump-and-treat methods
for water in the saturated zone. These methods are often impractical, which has led to the
ongoing development of supplemental and alternative methods, described briefly in
Chapter 6.
Wetlands have been used successfully for treating domestic, pharmaceutical and mining
related wastewater. There has been recent interest in using constructed wetlands to treat
arsenic laden waste. In a wetland, arsenic may exist in either reduced or oxidized forms, as a
component of organic arsenic – containing species and as insoluble sulfide minerals.
Research has been done to characterize the fate and transport of arsenic in soils; however,
little has been done to characterize these mechanisms for wetlands, a step which is necessary
in creating an efficient wetland treatment system. The research presented in Chapter 7
describes the first phase in modeling the fate of arsenic within a wetland and focuses on soil
characterization and estimation of the design life time using HYDRUS – 2D, a variably
saturated flow and contaminant transport model. A mixture of biosolids and sediments was
selected as the soil to be used in the proposed constructed wetland. The hydraulic
conductivity and adsorption coefficient for the soil were estimated using a soil – packed
vertical column. Four different scenarios were simulated: (i) top inlet and top outlet, (ii) top
inlet and bottom outlet, (iii) bottom inlet and top outlet, and (iv) bottom inlet top outlet with a
gravel layer at the bottom. The simulations were carried out for a bench scale (12 x 12x 54
cm) wetland but the results were extrapolated to a field - scale wetland. Simulations showed
that a design scenario of a bottom inlet with top outlet (with or without gravel layer at the
bottom) has the highest delay for breakthrough of arsenic in the outlet. Other avenues for
arsenic removal including plant uptake, and arsenic speciation and precipitation were not
included in the current simulation due to limitations of the software, but are to be added as the
research project continues.
The Emilia-Romagna Adriatic coast represents a strategic economic asset for Italy
because of its important environmental, historical and tourist value. An efficient
environmental protection and water resource management in this coastal zone is difficult
xii Luka F. König and Jonas L. Weiss

because of the many events that during the last century caused the ingression of large
volumes of brackish and saline water in the coastal aquifer.
River mouth enlargement to accommodate for tourist marinas has caused extensive
landward saltwater encroachment along rivers and canals. Tourist establishments along the
coastline have destroyed the original sand dunes continuity and their natural barrier effect to
beach erosion and saltwater intrusion. Strong natural and artificial subsidence, induced by gas
winning and deep groundwater exploitation, caused most of this territory to drop below mean
sea level and have modified the river regime and the normal groundwater flow.
In this situation, a drainage system is necessary to lower the water table level and keep
the land dry, so that trees roots stay above the saturated zone of the aquifer. The drainage
system management, however, has not accounted for the proximity to the sea. The result is an
unstable phreatic aquifer that is not able to contrast saltwater intrusion and is not beneficial to
maintaining healthy pine forests and avoiding soil salinization.
Water table and surface isosalinity maps have been created by monthly monitoring data.
These data show watertables below sea level and groundwater salinization; well testing and
geo-electric surveys provided a detailed lithological characterization and the depth of the
brackish-freshwater interface.
Analytical and numerical modeling (WATBAL, SUTRA, MOCDENS3D, and
SEAWAT) were used to calculate the water budget and study how past and present human
activities have affected the saltwater intrusion process and how the predicted future sea level
rise will further degrade the aquifer.
The feasibility and effects of some alternative actions such as controlled drainage,
freshwater storage in ponds along the coast, artificial recharge, river bottom sills,
optimization of the hydraulic infrastructure and coastal dune restoration have been examined.
In addition, the interactions between surface and ground waters as a connected resource
need to be further explored in order to find innovative strategies for coastal aquifer
management.
Finally, the authors stress in Chapter 8 how an integrated approach is necessary; history
has shown that a disaggregated water management, the present fragmentation of water
authorities and the lack of communications among different institutions and the stakeholders
are the major threats to the coastal environment and its water resources.
As presented in Chapter 9, the FVG Plain has been formed by river systems (mainly
Tagliamento and Isonzo) as well as deposition and reworking of marine and terrestrial
deposits. From a geomorphological and hydrogeological perspective, the FVG Plain consists
of two provinces separated by a resurgence belt that covers a strip 2 to 8 km wide and 80 km
long. The Upper Friuli Plain, composed mostly of calcareous and dolomitic gravels and
characterised by the lack of surface drainage waters hosts an unconfined aquifer (up to 150 m
thick). The Lower Friulian Plain is characterised by multi-layer confined aquifers. A detailed
reconstruction of the hydrogeological setting has indicated that there are a maximum of ten
confined aquifers in the southwestern part of the FVG Plain, whereas only 6 confined aquifers
are recognised in the southeastern part due to tectonic thrusts. However, considering their
chemical and physical characteristics these groundwater layers are subdivided in shallow and
deep confined aquifers. The shallow and deep confined aquifers have groundwater in the
temperature range of 10 to 20°C. The shallow and deep confined aquifers are separated by a
10 to 15 m thick impermeable layer of silty material at approximately 100 to 110 m depth.
The thickness of the multi-layered confined aquifer increases towards the Adriatic Sea where
Preface xiii

it reaches a thickness of up to 500 m. Thermal groundwaters (up to 60°C) have been found in
the southern part of the FVG Plain at depths of 350 m.
The pattern of geochemical and stable isotope variations suggests that the unconfined and
shallow confined groundwaters are recharged mostly by rainfall and local river (mainly
Tagliamento) infiltrations. This fast recharge process makes these groundwaters susceptible
to contamination to contamination. Four hydrogeological provinces have been recognised for
these subsurface groundwaters. Radiocarbon dating indicates that there is very little
continuity between the subsurface groundwaters and the deeper aquifers that have complex
groundwater circulations and have substantially changed during the varying temperature
regimes of the Holocene-Pleistocene. Significant late Quaternary sea-level fluctuations,
associated with alternating cooler and wetter periods, would have changed the hydraulic
gradients and partially or completely disconnected the deeper parts of the aquifer systems
from the more active surface circulations. Comparison with deep confined aquifers in other
regions of the Padain Plain indicates that the recharge rates of these deep confined aquifers
are low and that, consequently, the deep groundwaters are very sensible to overexploitation.
Our motivation to write Chapter 10 came from a real-life model simulating wastedisposal
embedded in an overlying rock. The main problem for our model is the large time-periods
necessary to obtain a realistic scenario to foresee the contamination process. Because of this
multi-scale problem, which occurred due to the coupled reaction terms of our underlying
system of convection-diffusion-dispersion-reactionequations, standard solver methods are
mostly ineffective to achieve realistic timeperiods. Due to this fact, the authors developed
analytical and numerical methods that allowed a computation over a large simulation period
of more than ten thousand years. The authors constructed discretization methods of a higher-
order with embedded analytical solutions, which allow large-time-steps without loss of
accuracy. Based on spatialand time-decomposition methods, the authors decouple complex
equations into simpler equations and use adequate methods to solve each equation separately.
For the explicit parts that are the convection-reaction-equations they use finite-volume
methods based on flux-methods with embedded analytical solutions. Whereas for the implicit
parts that are the diffusion-dispersion-equations they use finite-volume methods with central
discretizations. The authors analyze the splitting-error and the discretization error for our
methods. The main contribution of the paper is the design of analytical solutions with general
initial conditions that can be used to discretize the explicit parts with mass-conserved higher-
order methods. The verification of the new methods is done for benchmark and real-life
applications with our programme tool R3T. The testexamples and benchmark problems are
taken into account for the general initial conditions and verify our discretization- and solver-
methods with respect to the physical behavior. Based on the verification of realistic model-
problems the authors discuss a wastedisposal in three dimensions (3D) with large decay-
chains reacted and transported in a porous media with underlying flowing groundwater. For
the prediction of possible waste-disposal a computation with different located waste-locations
is discussed.
Chapter 11 presents some thoughts on the overall accuracy of the combination of flow
simulation and optimization models, which are used to optimize management of groundwater
resources.
Approximations introduced at different stages during the construction, or selection, of
groundwater flow and mass transport models are discussed first, together with their effect on
computational volume. Then the role of the optimization tool is presented, since the best
xiv Luka F. König and Jonas L. Weiss

choice of model for a flow simulation problem, when considered alone, might not be that
good in the frame of an optimization procedure, when the respective calculations should be
repeated many times. This case arises, when evolutionary methods are used as the
optimization tool. To clarify this point, the method of genetic algorithms, the most popular
evolutionary technique, is briefly outlined.
Finally the paper focuses on the integration of simplified flow (and mass transport)
simulation models in the optimization procedure, which is illustrated through a number of
examples. In the first, a simple conceptual model for optimization of coastal aquifer
management is presented. In the second an approximate form of the method of images is
introduced, which offers analytical calculation of hydraulic head drawdowns at the expense of
relaxing observation of the boundary conditions. Finally, in the last two examples, ways to
simplify simulation models of mass transport are discussed.
In order to manage groundwater supplies effectively, it is vital that the authors gain a
comprehensive understanding of the hydrological, physical and geological properties of the
subsurface at the highest possible spatial and temporal accuracies. Non-invasive geophysical
investigation techniques, such as electrical resistivity tomography (ERT), electromagnetic
conductivity (EM), Induced Polarisation (IP), self potential methods (SP) and Ground-
Penetrating Radar (GPR) have been used as successful groundwater characterisation tools
with Ground-Penetrating Radar becoming particularly popular in the past few years. As a
near-surface technique, GPR is unrivalled in its ability to map the three-dimensional structure
of the subsurface and can provide detailed, high-resolution information on the geology,
hydro-geological stratigraphy, water content and presence of preferential fluid pathways,
particularly in the vadoze-to-saturated zone. Recent studies have shown that the
spatial/temporal variation in GPR signal attenuation and velocity can provide important
additional information on the electrical properties of the sub-surface materials that, in turn,
can be used to assess the physical and hydrological nature of the groundwaters and identify
any areas of likely pore fluid contamination. In Chapter 12, the application of GPR for
groundwater and contaminant based studies will be discussed in terms of its historical
development, use, current application, good practice, practical limitations, advanced methods
of investigation and the future of groundwater-related GPR research. Most importantly, the
article will highlight how meaningful hydrological interpretations (such as fluid properties,
saturation index and the identification of hydrological pathways) can be extracted from the
GPR data as long as appropriate analysis techniques are used.
The problem of a coastal aquifers forced by oscillations in an adjacent sea and/or estuary
across a sloping boundary has recently received considerable theoretical attention. Despite
such a wealth of mathematical advancements, stringent testing of the limitations of these
models has yet to be undertaken. In all of the currently available analytical solutions it has
been assumed that a single length scale is sufficient to account for both the amplitude decay
rate and the rate of increase in phase lag (the wave speed) as the water table wave propagates
landward. All of the available field and laboratory data however indicate that this is not the
case. That is, the real part of the water table wave number (the amplitude decay rate) is not
equal to the imaginary part (the rate of increase in the phase lag). In Chapter 13, the detailed
laboratory measurements of Cartwright et al. [2004] are used to highlight the limitation of
assuming a single length scale in these mathematical models. In a step towards overcoming
this limitation, a new approximate analytical solution is derived which allows for two
different length scales as observed in the available data. In the absence of the ability to
Preface xv

accurately predict the water table wave number using basic aquifer parameters, all of the
solutions are applied to the data using water table wave numbers estimated from the data.
Accurately predicting the water table wave number based on measurable aquifer parameters
remains a challenge.
As explained in Chapter 14, available chemical and biogeochemical data were used in
developing a conceptual model to explain the source and distribution of the nitrogen
compounds in the Kuwait Group and the Dammam Formation, the two main aquifers in
Kuwait. Based on the available data and information, the anthropogenic contribution at sites
either inside or outside the political boundary of Kuwait as the main source of nitrate in the
groundwater of Kuwait has been discounted. Decay of organic matter or atmospheric nitrogen
fixation also appears to be an unlikely source of nitrate for the groundwater of Kuwait. Based
on the research work carried out at similar arid environment in other parts of the globe, it has
been hypothesized that nitrogen bearing minerals like nitratine (NaNO3), nitrocalcite
[Ca(NO3)2.4H2O] and/or niter (KNO3) may be present in the certain aquifer zones that have
given rise to the high concentration (≥ 15 mg/L) of nitrate in the upper part of the Kuwait
Group aquifer and in localized places in the Dammam Formation aquifer. The presence of at
least two of these minerals (nitratine and niter) has been observed in the surface sediments,
but need to be proved in the aquifers, however, before this hypothesis is accepted as the true
model for explaining the distribution of nitrate concentration in the groundwater of Kuwait. In
the agricultural area of Abdally in north Kuwait, the application of fertilizers may be
responsible for the observed high concentration of nitrate in the top part of the aquifer in this
area.
SHORT COMMUNICATION
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 3-14 © 2008 Nova Science Publishers, Inc.

USE OF ARTIFICIAL NEURAL NETWORKS


FOR PREDICTING OF GROUNDWATER
CONTAMINATION

Goloka Behari Sahoo1 and Chittaranjan Ray2,3,*


1
Department of Civil and Environmental Engineering, University of California at Davis,
One Shields Avenue, Davis, CA 95616, USA
2
Department of Civil and Environmental Engineering & Water Resources Research
Center, University of Hawaii at Manoa, 2540 Dole Street, Honolulu, HI 96822, USA
3
Now at the Faculty of Engineering, University of Georgia, Athens, GA 30603, USA

Abstract
Artificial neural networks are empirical mathematical tools proven to represent complex
relationships of hydrological systems. Neural networks are increasingly being applied in
subsurface modeling where intricate physical processes and lack of detailed field data prevail.
Two types of ANN models: Back propagation neural network (BPNN) and radial basis
function neural network (RBFN) are examined to predict the pesticide contamination of
domestic wells. Because sample collection, analysis, and re-sampling are expensive, a large
dataset is not available for ANN use in this study. This study presents analyzes of raw data for
preparation of input subsets for ANN use. Thus, a clustering technique is used to divide the
whole dataset into three subsets: training, validating, and testing. The sensitivity analysis was
carried out by deleting one or more input variables from the input data set to measure the
importance of one variable over the other in terms of ANN prediction performance. It
provides a sense of the effect of each parameter on pesticide occurrence in a well. The well
depth, depth to aquifer material from land surface, and on-site pesticide storage are found to
be important parameters in pesticide occurrence in well.

Keywords: Ground water quality, micro-genetic algorithm, back propagation neural network,
radial basis neural network, self-organizing maps.

*
E-mail address: cray@hawaii.edu (Corresponding author)
4 Goloka Behari Sahoo and Chittaranjan Ray

1. Introduction
Over 95% of rural residents and 50% of the total population in the United States rely on
groundwater for their drinking water (Solley, et al., 1998; Barbash and Resek, 1996).
Pesticides have been detected in groundwater across the nation (USEPA, 1990; Kolpin, et al.,
1994; Barbash and Resek, 1996; Barbash, et al., 1999) and the majority of the chemicals
found were herbicides (Weber, et al., 1997). Synthetic organic pesticides are used to control
weeds, insects, and other organisms in a wide variety of agricultural and nonagricultural
settings. However, the likelihood of pesticide entering into groundwater and surface water
must be reduced so that future water quality problems are not encountered. Since, pesticide
leaching into ground water results from a combination of soil and pesticide properties as well
as from the effect of the driving forces such as rainfall and regional-scale changes in land use;
regulatory principles should be derived based on information collected on a regional scale.
Sampling of individual wells is expensive and time consuming. Periodic sampling is
needed to confirm if land use factors are contributing to pesticide occurrence in ground water.
However, re-sampling effort is also expensive. Therefore, prediction of vulnerability of a well
to pesticide contamination from the available information is often important for public health
and regulatory perceptive on focusing target areas for re-sampling (Ray and Klindworth,
2000). Selecting a number of wells representing the region helps reduce the re-sampling
efforts and costs. This is possible by classifying the cause-effect information of the wells
collected on a regional scale into some clusters and taking one/two well from each cluster as
representative of that cluster.
A number of solute transport and non-point-source (leaching) models are available to
predict the movement of chemicals from the land surface to ground water (Carsel et al., 1984;
USDA-ARS (RZQM), 1992; Knisel, 1993; Simunek, et. al, 1998). These models require
physical descriptions of the porous media, initial and boundary conditions for flow and
transport processes, and the reactions occurring between the pesticide and the porous matrix
for predictive calculations. However, none of these models has the ability to predict pesticide
concentration at a well site because of these models are incapable of considering the complex
interaction between the soils and the pesticides, heterogeneity in soil physical and chemical
properties, and the uncertainty in estimating regional to local flow and transport parameters.
Further, these models cannot estimate concentrations of pesticides at a well using some of the
easily measurable cause-effect input factors such as well depth, distance to crop land, season
of sample collection, age of well, distance to septic system, and distance to on-farm
agricultural mixing and loading facilities etc (see Table 1). Moreover, these input parameters
are space and time dependent and are, at the same time, undergoing complex interactions. The
above models use complex mathematical equations to describe the physics of flow and
transport processes, detail data of soil characteristics, weather data, pesticide application rates
and time, and modeling skill. Data collection is expensive and time consuming. Detailed soil
characterization data collection over the entire state is a huge task. On the other hand, it has
been demonstrated that artificial neural networks (ANN) are capable of predicting water
quality and quantity issues in complex systems with reasonable accuracy (Maier and Dandy,
1996; Ray and Klindworth 2000, Mishra, et al., 2004).
Use of artificial neural networks (ANN) is now gaining tremendous momentum for
prediction purposes in the field of sub-surface and surface water quality (Schleiter et al.,
Use of Artificial Neural Networks for Predicting of Groundwater Contamination 5

1999; Maier and Dandy, 1999; Ray and Klindworth, 2000; Karul et al., 2000; Lischeid, 2001;
Aguilera et al., 2001; Brodnjak-Vončina et al., 2002; Zhang et al., 2002; Mujumdar and
Sasikumar, 2002) because of their specific features such as non-linearity, adaptivity (i.e.
learning from inputs parameters), generalization, and model independence (no a priori model
needed). The objectives of this paper are to (1) divide the available dataset into three subsets
using a clustering technique for ANN, (2) examine the applicability of ANNs to predict the
vulnerability of shallow ground water (from monitoring wells) to pesticide contamination
from the available data, and (3) to determine the influential parameters on pesticide
contamination in well waters through sensitive analysis.

2. Methodology
The ANN model requires three subsets of data: (1) training, (2) testing, and (3) validating.
These subsets must represent the same population for the ANN to achieve adequate
generalization ability (Bowden, et al., 2002). Therefore, Maier and Dandy (2000) proposed to
divide the original dataset into the three subsets using a trial-and-error method for achieving
optimum ANN performance. However, if 100 samples are divided into training, testing, and
validating subsets consisting of 50, 25, and 25 samples, respectively, there will be 100! / (50!
× 25! ×25!) = 1.3 × 1043 ways of arranging the samples. It would be practically impossible to
examine all the combinations. Thus, a self-organizing map (SOM) clustering technique
(Kohonen, 1982) is used in this study to group information of wells into groups of similar
characteristics and select wells from each groups for the three subsets. The details of SOM
clustering are described in Sahoo and Ray (2008).
Two ANN models: Radial basis function network (RBFN) and back propagation neural
network (BPNN) are examined to predict the pesticide contamination of domestic wells.
ASCE Task Committee (2000), Birikundavyi et al. (2002), Shi et al. (2005), Sahoo and Ray
(2006), and Alp and Cigizoglu (2007) stressed that ANN’s geometry and modeling
parameters have a significant influence on its performance efficiency and should be optimized
using a trial-and-error procedure. Since the solution space is large enough for each case, an
optimization technique, a micro-genetic algorithm, referred as μGA, (Carroll, 1999) is used in
this study (see Sahoo and Ray, 2008). The μGA optimized ANN model is referred to as μGA-
ANN in this study. Figure 1 describes the optimization of ANN geometry and modeling
parameters for three subsets of data prepared using SOM technique. The Subroutines
available in MATLAB version 7.1 (The Mathworks Inc., 2005) were modified and used to
create RBFN, BPNN and SOM models.
The performance efficiency of the network was estimated by comparing pesticide
concentration indices and ANN-estimated values. The ANN performances used in this study
are the correlation coefficient (R), mean error (ME), root mean square error (RMSE), and
mean square error (MSE). The mathematical expressions of R, ME, RMSE, and MSE are
described in Sahoo and Ray (2006). In brief, the ANN predictions are optimum if R, ME,
RMSE, and MSE are found to be close to 1, 0, 0, and 0, respectively. In the present study,
MSE is used only for the estimation of network training performance, whereas R, ME, and
RMSE are used to measure the prediction performance of ANN on the testing dataset, which
is independent of ANN network training and validating.
6 Goloka Behari Sahoo and Chittaranjan Ray

Figure 1. Flowchart for searching the three subsets of data having similar populations and for
optimization of the ANN’s geometry and modeling parameters.

3. Study Area
In this study, we used a dataset originally collected for the midcontinental United States
during 1991 to 1994 to estimate pesticide contamination in drinking water wells derived from
nonpoint sources, i.e., from pesticides applied to agricultural and nonagricultural fields
(Kolpin, et al., 1996). The data used in this study are the same as used in Mishra et al. (2004)
Use of Artificial Neural Networks for Predicting of Groundwater Contamination 7

and Sahoo and Ray (2008, in press). A summary of the data used is provided in Table 1. The
input data -- such as well depth, depth to aquifer material, and distance from well to cropland
-- do not hold any linear relationship between input parameters and observed pesticide
concentration; rather, the data are recognized in clusters. A generic procedure to utilize a set
of categorical input parameters for predictive purposes can be found in Ray and Klindworth
(2000), Brodnjak-Vončina, et al. (2002), Mishra, et al. (2004), and Sahoo, et al. (2005, 2006).
Using a similar approach, the measured values of all input parameters are categorized into
different classes, as shown in Table 1.

Table 1. Actual, Range-Specific, and Descriptive Input Parameters for the ANN Model

Actual /range-
Item specific/descriptive data Model
Parameter Type of data collected
no arranged in recognized clusters value
for ANN modeling
1 Well depth Actual depth (m) <7.6 4
7.6 – 15.2 3
>15.2 2
Unknown 1

2 Depth to aquifer material Actual depth (m) ≤1.5 4


>1.5 – 6.1 3
>6.1 – 15.2 2
>15.2 1

3 Age of well Actual year of excavation Year ≤ 1936 4


1936 < Year ≤ 1956 3
1956 < Year ≤ 1976 2
< 1976 1

4 Distance to cropland Actual distance or range- <6.1 4


specific data (m) 6.1 – 15.2 3
15.2 – 30.5 2
>30.5 1

5 Distance to barnyard Actual distance or range- <15.2 4


specific data (m) 15.2 – 30.5 3
30.5 – 61 2
> 61 1

6 Distance to septic systems Actual distance or range- <15.2 4


specific data (m) 15.2 – 30.5 3
30.5 – 61 2
> 61 1

7 Flush windows (time Actual days or range- <3 4


between pesticide specific data (days) 3 – 10 3
application and first 10 – 20 2
significant storm over 25 >20 1
mm per day or 12.5 mm per
day for two consecutive
days)
8 Goloka Behari Sahoo and Chittaranjan Ray

Table 1. Continued

Actual /range-
Item specific/descriptive data Model
Parameter Type of data collected
no arranged in recognized clusters value
for ANN modeling
8 Distance to streams or Rang-specific data (m) ≤30.5 2
other contaminant sources >30.5 1

9 Well-site topography Descriptive data Level land 4


Hill top 3
Depression 2
Hill slope 1

10 Season of sample collection Descriptive data Fall 4


Winter 3
Spring 2
Summer 1

11 Presence of irrigation well Descriptive data Yes 2


No 1

12 Spill or disposal site Descriptive data Yes 2


No 1

13 On-site pesticide storage Descriptive data Yes 2


No 1

14 Presence of animals Descriptive data Yes 2


No 1

15 Aquifer class Descriptive data Sand/gravel 2


bedrock 1

16 Pesticide leaching Actual concentration Real index value (0.1 to 10.0) Index
(μg/L) value

4. Analysis, Results and Discussions


4.1. Pre-processing of Input Data

Repetitive data (i.e., two or more identical input samples and each having a different target)
undermine ANN training (ASCE Task Committee, 2000). Because of the categorical inputs
for all 15 input parameters and pesticide concentration index as the target, a few repetitive
samples are found in the dataset. All repetitive samples are replaced by one input sample and
the average target value (i.e., the average pesticide concentration index) of all similar
samples, thus reducing the total available 631 samples to 572 samples. Moreover, the data are
scaled from 0 and 1 before presenting to ANN for training, validating, and testing.
Use of Artificial Neural Networks for Predicting of Groundwater Contamination 9

4.2. SOM Data Clustering

All 572 samples are presented to SOM that clusters the samples into groups according to their
similarly in characteristics for the occurrence of pesticide concentration. To examine the
number of clusters for higher/lower number of the output layer neurons, three SOM sizes: 64-
(i.e., 8 × 8 grid), 100- (i.e., 10 × 10 grid), and 144- (i.e., 1 2 × 12 grid) neurons are examined.
Interestingly, the SOM with 64-, 100-, and 144-neurons clustered all 572 samples into 58, 91,
and 132 groups, respectively. The reason for obtaining different clusters for three different
SOMs is - samples are uniquely different to each other based on their effects (i.e.,
characteristics) on pesticide concentration. Samples having similar effects on pesticide
concentration are clustered in one neuron. Thus, a higher size SOM produces more clusters,
narrowing down the differences in characteristics of samples in a cluster and vice versa.

4.3. Preparation of Three Subsets for μGA-ANN Models

The training, validating, and testing subsets are prepared from SOM-produced clusters. Two
samples from each cluster are selected, one for the training subset and another for the validating
subset, while other samples are placed in the testing subset. In the instance of a cluster containing
only one record, the record is placed in the training subset. However, if a cluster contains two
records, one record is placed in the training subset and the other in the validating subset. The
training dataset is checked to ensure that samples having the maximum and minimum pesticide
concentrations are present since ANN prediction is poor when it is used for extrapolation beyond
the range of data presented for training (ASCE Task Committee, 2000). Thus, the total numbers of
samples in the training subsets become 58, 92, and 132, for the 64-, 100-, and 144-neuron SOMs,
respectively. The validating subsets include 57, 85, and 116 samples while the testing subsets
include 457, 394, and 324 samples for the 64-, 100-, and 144-neuron SOMs, respectively.

4.4. Prediction Efficiency of μGA-ANN Models with Respect to Different


Training Subsets

The training, validating, and testing subsets are presented to the μGA-ANN models. The prediction
performance efficiency estimated using the testing subsets, which is not used in the ANN training,
are presented in Table 2. As can be seen, the prediction performance efficiency increases as the
number of samples in the training subset increases for the μGA-RBFN model. In case of μGA-
BPNN, the prediction performance efficiency for training subset containing 92 samples is highest
among other two cases. Therefore, to examine the case of more samples in the training subsets, the
third and fourth samples from each cluster are placed in the training and validating subsets,
respectively and the rests placed in the testing subset. If the number of samples available in a cluster
is less than the required amounts for the training and validating subsets, the first preference is given
to the training subset because training subset should contain information extending to the edges of
the modeling domain in all dimensions. The prediction performance efficiency of μGA-ANN
models using training subset containing two samples per cluster are presented in Table 3.
Interestingly, the prediction performance efficiency of μGA-ANN models using the 64-neuron
clustering data and two samples from each cluster for training subset outperforms to all other cases.
Table 2. Number of Samples in Training, Validating, and Testing Subsets and the Highest Performance Efficiency Values of μGA–
Optimized RBFN and BPNN

Datasets Performance efficiency of Performance efficiency of


No of samples in each subsets
prepared μGA-RBFN μGA-BPNN
scenarios Training Validating Testing R RMSE ME R RMSE ME
8 × 8 SOM 58 57 457 0.963 0.266 0.063 0.983 0.170 0.020

10 × 10 SOM 92 85 395 0.971 0.208 -0.003 0.987 0.140 -0.053

12 × 12 SOM 132 116 324 0.974 0.191 0.016 0.985 0.143 -0.002

Table 3. Number of Samples in Training, Validating, and Testing Subsets and the Highest Performance Efficiency Values of μGA–
Optimized RBFN and BPNN. The Bold Texts Represent Highest ANN Predictive Performance Efficiency among All Training Datasets.
Superscript 1 represents one increment, respectively. Bold texts represent the highest predictive efficiency among all cases

Datasets Performance efficiency of Performance efficiency of


No of samples in each subsets
prepared μGA-RBFN μGA-BPNN
scenarios Training Validating Testing R RMSE ME R RMSE ME
8 × 8 SOM1 113 107 352 0.979 0.188 0.027 0.991 0.127 0.001

10 × 10 SOM1 173 154 245 0.965 0.193 0.016 0.987 0.138 0.026

12 × 12 SOM1 228 186 158 0.956 0.153 0.008 0.986 0.090 0.005
Use of Artificial Neural Networks for Predicting of Groundwater Contamination 11

The primary reason for obtaining such results – 64-neuron SOM produced clusters most
of them having samples more than 4. So, there is equal distribution of samples among training
and validating subsets. However, 100- and 144-neruon SOM produced clusters, many of
which have samples less than 2. Thus, there are clear discrepancies of information in the three
subsets, particularly in the training and validating subsets. Although the training subset
contains samples from all clusters, the validating subset is void of samples from clusters
representing one or two samples. The validating subset, intended to prevent overfitting or
undertraining, could not help training the network adequately. Therefore, for a case where the
whole dataset does not contain enough samples, the SOM size should be selected by trial-and-
error process that produce no or a few clusters having one or two samples. This will also
facilitate in selecting more than one sample per cluster for training and validating subsets.

4.5. Sensitivity Analysis of Input Parameters on μGA-ANN Models


Prediction Efficiency

Sensitivity analysis of potential input factors contributing to well contamination such as well
depth, depth to aquifer material etc listed in Table 1 was carried out by deleting one input
variable from the input data set to measure the importance of that variable over others. Since
μGA-BPNN produced highest predictive performance efficiency using two samples per
cluster (see Table 3), μGA-BPNN with two samples for cluster training subset was used to
carry out the sensitivity analysis. The results presented in Table 4 show that the predictive
performance efficiency is lower for parameters: (1) well depth (R = 0.9893), (2) distance to
aquifer material (R = 0.9893), and (3) on-site pesticide storage facility (R = 0.9892) excluded
individually from the subsets. Moreover, predictive performance is highest when all
parameters are in the subsets. These results are consistent with Mishra et al. (2004) and Sahoo
et al. (2005). This infers that wells having ground water tables at shallow depths are more
vulnerable to pesticide contamination and vice versa. Similarly, wells located close to
pesticide storage facility within the farmer’s property are more vulnerable.

Table 4. Performance of the μGA-BPNN using two samples per cluster for training
subset. The bold texts represent the base case which includes all parameters
as shown in Table 1

Scenarios: Parameter excluded from the Predictive performance efficiency


input subsets R RMSE ME
Base case (no exclusion) 0.9910 0.1270 0.0010
Well depth 0.9893 0.1403 -0.0145
Depth to Aquifer material 0.9893 0.1360 0.0028
Age of well 0.9895 0.1330 -0.0144
Distance to cropland 0.9897 0.1362 -0.0209
Distance to barnyard 0.9895 0.1774 -0.0196
Distance to septic system 0.9898 0.1309 0.0065
12 Goloka Behari Sahoo and Chittaranjan Ray

Table 4. Continued

Scenarios: Parameter excluded from the Predictive performance efficiency


input subsets R RMSE ME
Flush window 0.9896 0.1398 -0.0071
Well-site topography 0.9897 0.1339 0.0008
On site pesticide storage 0.9892 0.1353 0.0058
Well site topography 0.9897 0.1339 0.0008

5. Conclusion
Prediction of pesticide occurrence in rural domestic wells is important. However, it is difficult
to get information on soil characteristics and underlying physics of the pesticide movement in
subsoil on a regional scale. Therefore, empirical models such as artificial neural networks are
commonly employed as an alternative to predict the pesticide occurrence in a well using the
available cause-effect information. Also, it is often difficult to get a large dataset for ANN.
Since sampling and re-sampling is time-consuming and expensive, classifying the cause-
effect information of the wells collected on a regional scale into some clusters and taking
one/two well from each cluster as representative of that cluster is important. In this study,
self-organizing map is used to cluster the wells into groups of similar characteristics.
Two ANN models: (a) BPNN and (b) RBFN were optimized using μGA and were used
to predict pesticide occurrence in a well. Results showed that selection of a suitable number
for SOM neurons is important because a small SOM groups samples of wider characteristics
into a cluster while a large SOM is unable to generate clusters each with enough samples to
contribute equally to the three subsets. The SOM size found appropriate was a 8 × 8 grid that
produced clusters each having more than 4 samples per clusters in this study. Given a suitable
size of SOM, μGA-ANN models using two samples per cluster in the training subsets
outperform to all other cases.
Sensitivity analysis was carried out to measure the importance of that input parameter
over other. This is done by deleting the input parameter from the subset. Results showed that
depth of well, depth to aquifer material from land surface, and on-site pesticide storage
facilities are important. These findings are consistent with previous studies.
The results presented herein are based on the available dataset and on one site. However,
the methodology can be applied to any other site with necessary modification.

Acknowledgments

Partial funding for this research was provided by the United States Department of
Agriculture, National Research Initiative/Competitive Grants Program (Contract Number 99-
35102-8551, Project Number HAWR 1999-01133).
Use of Artificial Neural Networks for Predicting of Groundwater Contamination 13

References
Aguilera, P.A., A.G. Frenich, J.A. Torres, H. Castro, J.L. M. Vidal, and M. Canton, (2001).
Application of the Kohonen neural network in coastal water management:
methodological development for the assessment and prediction of water quality. Water
Research, 35 (17), 4053-4062.
Alp, M. and H.K. Cigizoglu (2007). Suspended sediment load simulation by two artificial
neural network methods using hydrometeorological data, Environ. Modell. Softw., 22(1),
2–13.
ASCE Task Committee (2000). Artificial neural network in hydrology, J. Hydrol. Eng., 5(2),
124–144.
Barbash, J.E. and E.A. Resek (1996). Pesticides in ground water: distribution, trends, and
governing factors, 590 pp., Lewis Publishers, Boca Raton, FL.
Barbash, J.E., G.P. Thelin, D.W. Kolpin, and R.J. Gillom (1999). Distribution of major
herbicides in ground water of the United States, Water Resources Investigations Report
98–4245, 64 pp., U.S. Geological Survey.
Birikundavyi, S., R. Labib, H.T. Trung, and J. Rousselle (2002). Performance of neural
networks in daily streamflow forecasting, J. Hydrol. Eng., 7(5), 392–398.
Bowden, G.J., H.R. Maier and G.C. Dandy (2002). Optimal division of data for neural
networks models in water resources applications, Water Resour. Res., 38(2), doi:
10.1029/2001WR000266.
Brodnjak-Vončina, D., D. Dobčnik, M. Novič, and J. Zupan (2002). Chemometrics
Characterisation of the quality of river water, Anal. Chim. Acta, 462(1), 87–100.
Carroll, D.L. (1999). FORTRAN Genetic Algorithm (GA) Driver version 1.7.0, Available
from http://www.cuaerospace.com/carroll/ga.html (accessed during January 2006)
Carsel, R.F, C.N. Smith, L.A. Mulkey, J.D. Dean, and P. Jowsie (1984). Pesticide root zone
model (PRZM), Release 1, EPA–600/3–84–109, USEPA, Washington, D.C.
Karul, C., S. Soyupak, A.F. Cilesiz, N. Akbay, and E. Germen, (2000). Case studies on the
use of neural networks in eutrophication modeling. Ecological Modelling, 134, 145-152.
Knisel, W.G. (1993). GLEAMS: groundwater loading effects of agricultural management
systems, version 2.10, University of Georgia Coastal Plain Experimental Station, Tifton,
GA.
Kohonen, T. (1982). Self-organized formation of topologically correct feature maps, Biol.
Cybern., 43, 59–69.
Kolpin, D.W., E.M. Thurman and D.A. Goolsby (1996). Occurrence of selected pesticides
and their metabolites in near surface aquifers of the Midwestern United States, Environ.
Sci. Technol., 30, 335–350.
Kolpin, D.W., M.R. Burkart and E.M. Thurman (1994). Herbicides and nitrate in near-surface
aquifers in the mid-continental United States, 1991, Water Supply Paper 2413, U.S.
Geological Survey.
Lischeid, G. (2001). Investigating trends of hydrochemical time series of small catchments by
artificial neural netwroks, Physics and Chemistry of the Earth, 26 (1), 15-18.
Maier, H.R. and G.C. Dandy (2000). Neural networks for the prediction and forecasting of
water resources variables: a review of modelling issues and applications, Environ.
Modell. Softw., 15(1), 101–124.
14 Goloka Behari Sahoo and Chittaranjan Ray

Maier, H. R. and G.C. Dandy (1996). The use of artificial neural networks for the prediction
of water quality parameters, Water Resources Research, 32, 1013-1022.
Maier, H. R., G. C. Dandy (1999). Empirical comparison of various methods for training
feed-forward neural networks for salinity forecasting, Water Resources Research, 35 (8),
2591-2596.
Mishra, A., C. Ray and D.W. Kolpin (2004). Use of qualitative and quantitative information
in neural networks for assessing agricultural chemical contamination of domestic wells,
J. Hydrol. Eng., 9(6), 502–511.
Mujumdar, P.P. and K. Sasikumar (2002). A fuzzy risk approach for seasonal water quality
management of a river system. Water Resources Research, 38 (1), (5) 1-(5) 9.
Ray, C. and K.K. Klindworth (2000). Neural networks for agrichemical vulnerability
assessment of rural private wells, J. Hydrol. Eng., 5(2), 162–171.
Sahoo, G. B. and C. Ray (2006). Flow forecasting for a Hawaii stream using rating curves
and neural networks, J. Hydrol., 317, 63–80.
Sahoo, G.B., C. Ray and H.F. Wade (2005). Pesticide prediction in ground water in North
Carolina domestic wells using artificial neural network, J. Ecol. Modell., 183, 29–46.
Sahoo, G.B., C. Ray, E. Mehnert and D.A. Keefer (2006). Application of artificial neural
networks to assess pesticide contamination in shallow groundwater, Sci. Total Environ.,
367, 234–251.
Sahoo, G.B. and C. Ray (2008). Micro-genetic algorithms and artificial neural networks to
assess minimum data requirements for prediction of pesticide concentrations in shallow
groundwater on a regional scale. Water Resources Research, doi:10.1029/
2007WR005875.
Schleiter, I.M., D. Borchardt, R. Wagner, T. Dapper, K. Schmidt, H. Schmidt, and H. Werner,
(1999). Modelling water quality, bioindication and population dynamics in lotic
ecosystems using neural networks. Ecological Modelling, 120, 271-286.
Shi, D., D.S. Yeung and J.Gao (2005). Sensitivity analysis applied to the construction of
radial basis function networks, Neural Networks, 18(7), 951–957.
Simunek, J., M. Sejna and R. van Genuchten (1998). HYDRUS-1D manual of versions 2.0,
U.S. Salinity Laboratory, Riverside, CA.
Solley, W.B., R.R. Pierce and H. A. Perlman, (1998). Estimated use of water in the United
States in 1995, Circular, 1200, 71 pp., U.S. Geological Survey.
The MathWorks, Inc. (2005). MATLAB version 7.1, 3 Apple Hill Drive, Natick,
Massachusetts, USA.
USDA–ARS (1992). Root zone water quality model (RZQM) v. 1.0., Technical Documentation,
GSPR Report 2, USDA–ARS Great Plains Systems Research Unit, Fort Collins, CO.
USEPA (1990). National survey of pesticides in drinking water wells, phase I report, EPA
570/9–90–015, U. S. Environmental Protection Agency, Office of Water and Office of
Pesticides and Toxic Substances, Washington, D.C.
Weber, J.B., R.A. McLaughlin, H.F. Wade, E. Morey, and N.C. Raleigh (1997). Finding and
predicting pesticides in ground water in North Carolina, Consumer Environmental Issues:
Safety, Health, Chemicals and Textiles in the Near Environment, in International
Symposium Proceedings, pp. 211–220, St. Petersburg, FL.
Zhang, Y., J. Pulliainen, S. Koponen, and M. Hallikainen, (2002). Application of an empirical
neural network to surface water quality estimation in the Gulf of Finland using combined
optical data and microwave data. Remote Sensing of Environment, 81, 327-336.
RESEARCH AND REVIEW STUDIES
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 17-77 © 2008 Nova Science Publishers, Inc.

Chapter 1

GIS-BASED AQUIFER MODELING AND PLANNING


USING INTEGRATED GEOENVIRONMENTAL AND
CHEMICAL APPROACHES

Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios*


Technological Educational Institute of Crete, Department of Natural Resources and
Environment

Abstract
Groundwater (GW) is one of the most valuable natural resources and for that reason, the
GW protection and management is vital for human evolution, socio-economic development
and ecological diversity. During the last decades, the continuously increasing need of water
has led to a rapidly growing awareness in the field of GW management. At the same time over
exploitation and pollution of water resources are threatening the ecosystems. The combination
of these two problems which have acquired worldwide dimensions has forced many scientists
working in relative fields to search new, multidisciplinary approaches to address them.
Effective management and protection of groundwater resources require detail knowledge and
quantitative/qualitative characterization of aquifers. Thus, modeling and planning of the GW
through the use of modern technologies and approaches have become of high priority towards
this direction.
Despite advances in modeling tools and geoenvironmental and chemical methods, aquifer
characterization remains an extremely difficult problem due to spatial heterogeneity, temporal
variability and coupling between chemical, physical, and biological processes. Solution to this
problem involves multidimensional data integration. The concept of data fusion involves the
merging of multiple data types to develop more reliable predictive models and to address
basic and applied scientific questions concerning GW modeling.
With the advent of powerful computers and the advances in geoenvironmetal methods
and space technology, efficient techniques for water management have evolved. Remote
sensing (RS), geographic information system (GIS), innovative geophysical methods
(electrical, electromagnetic and seismic) and GW flow simulation codes are such techniques
of great significance. These techniques, with the aid of chemical analyses, have fundamentally
reassigned the ways to manage natural resources in general and water resources in particular.

*
E-mail address: soupios@chania.teicrete.gr , tel.: 00302821023037, fax: 00302821023042. Corresponding author:
Pantelis Soupios, 3 Romanou, Halepa, 73133, Hania, Crete, GREECE
18 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

The combined application of geophysical methods and chemical analysis of soil and
water samples can provide detailed information about aquifer hydraulic parameters and
possible contamination of the GW with the highest possible resolution. This information fully
integrated with other data derived either from RS (such as land use, land cover, tectonic,
bedrock feature) or existing data in map / time series / tabular form (e.g. hydro-
meteorological, borehole data) in a GIS environment can be used for GW modeling after the
application of a simulation code (such as MODFLOW). After the model has run, the results of
the simulation (estimation of GW potential and prediction of aquifer response to groundwater
pumping and recharge and the flow path analysis) can be exported into a GIS for post
processing (PP). The PP step allows for data layers to be developed and displayed in map
forms, allowing easy examination and interpretation of the results in their spatial context.
Finally, for each data layer, a weight relative to its importance can be defined and a decision
support system (DSS) for groundwater management can be evolved.
The overall results demonstrate that the application of GIS in conjunction with the
aforementioned geoenvironmental technologies and chemical analysis provide a powerful tool
to study groundwater resources and design a suitable exploration and management plan.

Introduction
Groundwater is globally important for human consumption as well as for the support of
habitat and for maintaining the quality of base flow to rivers. As it becomes naturally filtered
by percolation towards the aquifer, usually it is clear, colorless, and free from microbial
contamination requiring minimum treatment. A threat is now posed by an ever-increasing
number of soluble chemicals from several urban and industrial activities and from modern
agricultural practices (fertilizers, etc). Nevertheless, landslides, fires and other surface
processes that increase or decrease infiltration and/or expose or blanket rock and soil surfaces,
which interact with downward-moving surface water, may also affect the quality of shallow
groundwater (Babiker et al. 2007).
As a consequence of poor water quality and limited water availability, there is clearly an
urgent need for rapid reconnaissance techniques that allow an assessment of groundwater
vulnerability over large areas, despite the fact that there may be only limited secondary data
(Al-Adamat et al. 2003).
Groundwater protection begins with the assessment of the sensitivity of its environment.
Various techniques and methodologies have been developed to evaluate environmental
impacts associated with groundwater pollution, among which, the concept of aquifer
vulnerability. The term “aquifer vulnerability” was introduced by Margat (1968) for
expressing the degree of protection that the natural environment provides against the ingress
of pollutants to the groundwater. Since then, several definitions of vulnerability have been
proposed (Vrba and Zaporozec 1994; Doerfliger et al. 1999; Gogu and Dassargues 2000;
Daly et al. 2002; Zwahlen 2004) without any standard definition of aquifer vulnerability.
According to Vrba and Zaporozec (1994), groundwater vulnerability represents the intrinsic
properties of aquifer systems as a function of their sensitivity to human and natural activities.
It can be defined as the possibility of percolation and diffusion of contaminants from the
ground surface into the groundwater system. The concept of groundwater vulnerability
includes two particular notions: intrinsic vulnerability and specific vulnerability. Intrinsic
vulnerability refers to the vulnerability of groundwater to contaminants generated by human
activities taking into account the inherent geological, hydrological and hydrogeological
characteristics of an area but being independent of the nature of the contaminants and
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 19

pollutant attenuation. Specific vulnerability is used to define the vulnerability of groundwater


to particular contaminants or a group of contaminants taking into account the contaminant
properties and their relationship with the various components of intrinsic vulnerability
(Doerfliger et al. 1999; Gogu and Dassargues 2000). Groundwater vulnerability to
contamination can be defined as the propensity or likelihood for contaminants to reach some
specific position in the groundwater system after their introduction at some point above the
top of the uppermost aquifer (Rao and Alley 1993). Vulnerability is usually considered as an
intrinsic property of a groundwater system that depends on its sensitivity to human and/or
natural impacts. Specific or integrated vulnerability, on the other hand, combines intrinsic
vulnerability with the risk of the groundwater being exposed to the loading of pollutants from
certain sources (Vrba and Zaporozec, 1994).
In general, the level of groundwater contamination is determined by the natural
attenuation processes occurring within the zone between the pollution source and the aquifer.
The concept of assessing groundwater vulnerability and contamination risk is based on an
origin-pathway-target model. Origin is the term used to describe the location of a potential
contaminant release. The pathway comprises the passage of contaminants from the origin to
the target, i.e. the water that shall be protected. Resource protection aims to protect the whole
aquifer; source protection aims to protect a spring or well. For resource protection, the
groundwater surface is defined as the target, and the pathway consists of the unsaturated zone.
For source protection, the pathway additionally includes the flow in the aquifer towards the
spring or well (Goldscheider 2004). From a quantitative point of view, three aspects are
important for vulnerability assessment: the travel time of a contaminant from the origin to the
target, the attenuation along its pathway, and the duration of a contamination at the target
(Brouyère 2004). This approach makes it possible to validate vulnerability assessments by
means of artificial tracer tests, and chemical and microbiological groundwater quality data
(Goldscheider et al. 2001; Holman et al. 2005; Perrin et al. 2004).
Groundwater vulnerability mapping is based on the idea that some land areas are more
vulnerable to groundwater contamination than others (Piscopo 2001) and, since the 1980s,
basic vulnerability indices have been developed extensively for planning purposes in many
areas of the world (e.g. Carter et al. 1987; NRA 1994). Planning is often based on concepts of
both resource and source protection, although the success of EPA in U.S. or EU and national
government policies on groundwater protection have often been brought into question since
general policies often require significant local modification and improved dissemination of
information (e.g. Foster and Ilbery 1992; Foster and Thorn 1993). The vulnerability concept
is implemented by classifying a geographical area with regard to its susceptibility to
groundwater contamination rather than using dynamic groundwater models, because
groundwater models often require data which is unavailable in many parts of the world (Knox
et al. 1993).
The present chapter highlights geographic information system (GIS) and remote sensing
(RS) technologies and presents a state of the art review on the application of these two
emerging techniques coupled with geochemical and geophysical approaches in a river basin
in Greece for groundwater modeling/management and planning.
20 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Defining Geographic Information System (GIS) and Remote


Sensing (RS)
Geographic Information System is defined as a computer-assisted mapping and cartographic
application, a set of spatial-analytical tools, a type of database systems, or a field of academic
study (Lo and Yeung 2003). In order to provide a simple working definition of GIS, the two
widely used definitions are: (i) “GIS is a system of hardware, software, and procedures
designed to support the capture, management, manipulation, analysis, modeling, and display
of spatially referenced data for solving complex planning and management problems” (Rhind
1989); and (ii) “GIS is a computer system capable of assembling, storing, manipulating, and
displaying geographically referenced information” (USGS 1997). The basic ideas contained
in these two definitions have been adopted in GIS textbooks such as by Burrough (1986),
Aronoff (1989), DeMers (2000) and Clarke (2001). Simply put, GIS is a set of computer-
based systems for managing geographic data (i.e., spatial data having the reference to
geographic space and the representation at geographic scale) and using these data to solve
various spatial problems, while the skills and procedures for collecting, managing, and using
geographic information entails a comprehensive body of scientific knowledge from which
these skills and procedures are developed (Lo and Yeung 2003).
Remote sensing (RS) can be defined as the observation of targets or processes from a
distance, without being in a physical contact with the target. This term usually refers to the
gathering and processing of information about earth’s environment, particularly its natural
and cultural resources, through the use of photographs and related data acquired from an
aircraft or a satellite (Simonett 1983). Thus, remote sensing also includes data analysis which
involves the methods and processes for extracting meaningful spatial information from the
remotely sensed data for direct input to a geographic information system. In RS, conventional
aerial photography and satellite remote sensing instruments that obtain pictures of visible,
near-infrared (NIR) and thermal infrared (TIR) energy belong to passive remote sensing
techniques, while the radar and lidar belong to active remote sensing techniques.

Basic Concept of GIS

The geographic data can be represented in GIS as objects or fields. In the object approach,
real-world features are represented by simple objects such as points, lines and polygons. The
objects (representing features) are characterized by geometry, topology, and non-spatial
attribute values. On the other hand, in the field approach, real-world features are represented
as fields of attribute data without defining objects. This approach provides attribute values in
any location. In GIS, the distinction between objects and fields is associated with vector data
models and raster data models (Goodchild 1992). The vector data model is an object-based
approach for representing real-world features and is best used to represent discrete objects.
All vector data models are built on two common and interrelated concepts: the decomposition
of spatial objects into basic graphical elements, and the use of topology (spatial relationship)
and geometry (coordinates) to represent spatial objects.
The raster data model is a field-based approach for representing real-world features and is
best employed to represent continuous geographic phenomena. This model is characterized by
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 21

sub-dividing a geographic space into grid cells with values being assigned to each cell. The
linear dimensions of each cell define the spatial resolution of the data, which is determined by
the size of the smallest object in the geographic space to be represented. This size is also
known as the “minimum mapping unit (MMU)”. In raster data models, each cell is usually
restricted to a single value. Hence, multiple layers are needed to represent the spatial
distribution of a number of parameters (variables). A raster-based GIS has advantages over a
vector-based GIS because virtually all types of data including attribute data, image data,
scanned maps, and digital terrain models can be represented in raster form (Van Der Laan
1992). Also, the vector data model is conceptually more complex and more technically
difficult to implement than the raster data model. However, which data format is better to be
used depends actually on the type of applications.
Data analysis tools include aggregation, classification, measurement, overlay, buffering,
networks, and map algebra. Aggregation helps the user in interpreting the data, classification
allows the user to classify areas within a map, and measurement is used to determine the size
of any area. The overlay function allows the user to "stack" map layers on one another.
Buffering examines an area that surrounds a feature of interest such as a point. Network
functions examine the movement of objects along an interconnected pathway (e.g., traffic
flow along a map of highway segments). Map algebra utilities allow the user to specify
mathematical relationships between map layers.

Data Extraction from Satellite Imagery

Remote sensing with its advantages of spatial, spectral and temporal availability of data
covering large and inaccessible areas within short time has become a very important tool in
exploring, evaluating, and managing vital groundwater resources (Chowdhary et al. 2003).
The hydrogeologic interpretation of satellite data have been proved to be a valuable survey
tool in areas of the world where little geologic and cartographic information exists or the
existing information is not accurate (Engman and Gurney 1991).
Specifically, satellite data provide quick and useful baseline information about the factors
controlling the occurrence and movement of groundwater like geology, lithology,
geomorphology, soils, land use/cover, drainage patterns, lineaments, etc. (Bobba et al. 1992;
Meijerink 2000). However, all the controlling factors have rarely been studied together
because of the non-availability of data, integrating tools and/or modeling techniques.
Structural features such as faults, fracture traces and other such linear or curvilinear features
can indicate the possible presence of groundwater (Engman and Gurney 1991). Similarly,
other features like sedimentary strata (i.e., alluvial deposits and glacial moraines) or certain
rock outcrops may indicate potential aquifers. Shallow groundwater could also be inferred by
soil moisture measurements and by changes in vegetation types and patterns (Nefedov and
Popova 1972). In arid regions, vegetation characteristics may indicate groundwater depth and
quality. Groundwater recharge and discharge areas in drainage basins can be detected from
soil types, vegetation, and shallow/perched groundwater (Todd 1980). Furthermore,
differences in surface temperature (resulting from near-surface groundwater) measured by
remote sensing sensors have also been used to identify alluvial deposits, shallow
groundwater, and springs or seepages (Myers and Moore 1972; Heilman and Moore 1981;
van de Griend et al. 1985).
22 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Once the required satellite images are purchased in digital form, the next step is to
process the images for extracting the desired spatial and thematic information; satellite
images without processing are not of much use, especially for scientific studies. This complex
processing, known as digital image processing, is done with the help of a computer by using
image processing software packages (Lillesand and Kiefer 2000). Figure 1illustrates the
major steps for extracting data from digital satellite images. Clearly, several operations are
needed for extracting the required data and/or information. Since in most cases the data
obtained from satellite systems are input to a GIS for analyses, modeling and preparation of
thematic maps, a conversion from raster data to vector data is necessary. Most digital image
processing software and advanced GIS software packages can perform raster to vector
conversion. This conversion, however, will not be required if the GIS technology is able to
process digital images and handle both vector and raster data (Lo and Yeung 2003).

Satellite Imagery
Data Extraction

Step1 Step 2 Step 3

Image Image
Acquisition of Pre-Processing Interpretation
Image Raster to Vector
Satellite Images (Atmospheric & Input to GIS
Enhancement conversion
Geometric Extraction of
Corrections) thematic maps

Figure 1. The major three steps for extracting data from digital satellite images.

In a first pre-processing step the satellite images and the Digital Elevation Model (DEM)
are geocoded, using the corresponding topographic maps of the study area. For geocoding,
input data have to be preprocessed, for example, adjacent map sheets have to be scanned and
joined. The appropriate raster layers (satellite data, DEM) are overlaid and manipulated, in
order to derive useful preliminary conclusions concerning the localization of the desired
information. In the case of lineaments which are indicators of possible faulting or existence of
other tectonic features, lineament- enhancement techniques (e.g. Rigol and Chica-Olmo 1998)
can be applied on the digital datasets, in order to investigate the basic tectonic pattern of the
study area.
Application of spatial filtering techniques on the satellite images can include: (a) Linear
and nonlinear convolution filters for edge enhancement (directional masks, linear gradient
filters, edge-enhancement sun-angle filters, high-pass filters) (Argialas et al. 1988; Biedny
and Monroy 1991; Russ 1992; Jensen 1996; Gupta 1991); and (b) Edge-detection filters
(Haralick 1984; Jain 1989; Pratt 1991; Jensen 1996; Rokos et al. 2000). The enhanced images
are expected to contribute to the interpretation of major tectonic features (faults, fracture
zones), geomorphologic structures and certain drainage network segments.
In the next data-processing phase, a vector layer of lineaments will be created by screen
digitizing on the enhanced images and/or on available high resolution images. High resolution
images usually provide a significant amount of information for the tectonic structure of the
research area, because their high spatial resolution contributes to the enhancement of the
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 23

weaker lineaments, which may not be apparent on the DEM. Besides the topographic
lineaments, lineaments resulting from vegetation alignment can also be detected on high
spatial resolution images. Furthermore, such images provide more segments (branches) of the
existing drainage network in comparison to the shaded relief maps, and all expected tectonic
trends (major and minor) of the study area can be delineated.
The derived vector (lineament maps) and raster (satellite images and DEM) layers have
to be implemented in a GIS environment. The lineament map and the processed images can
be manipulated in a GIS environment in order to study the relation of the tectonic pattern to
both the hydrogeology and the geomorphology of the study area.
Numerous other techniques can be applied depending on the desired information such as
vegetation indices and classification algorithms for land-use and geological maps, soil indices
for soil maps, or thermal infrared bands Digital Numbers to temperature based on radiance
value conversion.
A table with detailed information about the important physical features of the landscape
which can be derived from satellite imagery or aerial photographs and used for assessing
groundwater conditions can be found in Jha et al. (2007). Excellent reviews of remote sensing
applications in groundwater hydrology are presented in Farnsworth et al. (1984), Waters et al.
(1990), Engman and Gurney (1991), Meijerink (2000) and Jha et al. (2007). These reviews
indicate that remote sensing has been widely used as a tool, mostly to complement standard
geophysical techniques. Meijerink (2000) recognizes the value of remote sensing in
groundwater recharge-based studies and suggests that it can significantly aid to the
conventional assessment and modeling techniques.

GIS and RS Applications in Groundwater Modeling and


Management
Since Margat (1968) and Albinet and Margat (1970) introduced the concept of the
vulnerability of groundwater to contamination, the international scientific community has
shown an increasing interest in groundwater protection, applying modern computer-based
tools such as the GIS environment (Foster and Hirata 1988; Adams and Foster 1992; Drew
and Hotzl 1999; Zwahlen 2004).
The geographic information system (GIS) has emerged as an effective tool for handling
spatial data and decision making in several areas including engineering and environmental
fields (Stafford 1991; Goodchild 1993; Bonham-Carter 1996; Faust et al. 1991; Hinton 1996).
Moreover, the combined use of remote sensing and GIS is a valuable tool for the analysis of
voluminous hydrogeologic data and for the simulation modeling of complex subsurface flow
and transportation processes under saturated and unsaturated conditions (e.g. Watkins et al.
1996; Loague and Corwin 1998; Gogu et al. 2001; Gossel et al. 2004).
In recent years, GIS methods have been widely used in groundwater vulnerability
mapping (Loague et al. 1996; Evans and Myers 1990; Hrkal 2001; Rupert 2001; Lake et al.
2003). The major advantages of GIS-based mapping are the best combination of data layers
and rapid change in the data parameters used in vulnerability classification. GIS are designed
to collect diverse spatial data in order to represent spatially variable phenomena by applying a
series of overlay analysis of data layers that are in spatial register (Bonham-Carter 1996). It is
24 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

concluded that the combined use of RS and GIS technologies has great potential to
revolutionize groundwater monitoring and management in the future by providing unique and
new data to supplement the conventional field data.
Many GIS functions currently available or under development could further meet the
requirements of process-based approaches for analyzing subsurface phenomena (Gogu et al.
2001). Undoubtedly, the GIS technology allows for swift organization, quantification, and
interpretation of a large volume of hydrologic and hydro-geologic data with computer
accuracy and minimal risk of human errors. Unlike surface water hydrology, the applications
of RS and GIS techniques in groundwater hydrology have received only cursory treatment
and are less documented. Furthermore, the roles of RS and GIS in groundwater hydrology
have been reported separately in the past and a combined treatise with comprehensive reviews
is not reported to date.
As far as the remote sensing applications in groundwater hydrology are concerned, aerial
photographs and visible and near-infrared satellite images have been used for groundwater
exploration experimentally since 1960s with only limited success (Engman and Gurney
1991). The absence of spectral resolution did not allow the effective use in groundwater
prospecting. However, with the development of high resolution multi-spectral satellite
sensors, the use of satellite imagery for groundwater prospecting dramatically increased in
late 1980s (Waters et al. 1990; Engman and Gurney 1991; Meijernik 2000; Jackson 2002).
Due to the high cost of drilling, the use of remote sensing techniques has been proved a very
cost-effective approach in prospecting and preliminary surveys. Generally, the analysis of
aerial photographs or satellite imagery is recommended prior to ground surveys and
fieldwork, because it may eliminate areas of potentially low water-bearing strata and may also
indicate promising areas for intensive field investigations (Revzon et al. 1983). It should be
noted however that the adoption of remote sensing does not eliminate the in situ data
collection, which is still essential to verify the accuracy and the interpretation of the remote
sensing data. Of course, remote sensing helps minimize the amount of field data collection.
Reviews of GIS and RS techniques, combined with other geoenvironmental applications
in hydrology and water management have been presented by several researchers during early
nineties and mid-nineties (Zhang et al. (1990), DeVantier and Feldman (1993), Ross and Tara
(1993), Schultz (1993), Deckers and Te Stroet (1996), and Tsihrintzis et al. (1996)). These
reviews indicate that these applications in hydrology and water management are essentially in
a modeling dominated context. Longley et al. (1998), on the other hand, while presenting the
development of geocomputation, discuss various geoscientific applications of GIS as well as
the role of geocomputation in the development and application of GIS technologies. Although
the use of GIS in groundwater modeling studies dates back to 1987, its use for surface-water
modeling has been more prevalent than for groundwater modeling because the available
standardized GIS coverages are primarily of the land surface; few standardized coverages of
hydrogeologic properties are available (Watkins et al. 1996). Watkins et al. (1996) presented
an excellent overview of GIS applications in groundwater-flow modeling and discussed its
usefulness and future directions. On the other hand, Pinder (2002) provided step-by-step
procedures for groundwater flow and transport modeling using GIS technology. The current
status of GIS applications in groundwater hydrology is presented in the following sections. A
detailed survey of the past literature concerning mainly the applications of GIS techniques in
studying groundwater problems is also presented. However, in the present chapter, the
applied RS-based groundwater studies are not included as we are focused mainly on GIS
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 25

applications. RS is mentioned as a dynamic source of hydrogeological data without giving


details for the processes of data extraction. Based on this comprehensive literature review, the
applications of GIS techniques in groundwater vulnerability mapping have been categorized
into two groups based on the type of aquifer (porous and karstic) and the studies pertaining to
each group are described in the following sections.

Groundwater Vulnerability Mapping - GVM

Vulnerability mapping is defined as a technique for quantifying the sensitivity of the resource
to its environment, and as a practical visualization tool for decision-making. Maps are
produced from a set of decisional criteria linked to a number of physical parameters
representing the study site; the choice depends on the model used. Vulnerability maps can be
calculated with the aid of a GIS which allows spatial data gathering and, at the same time,
gives a mean for data processing, such as geo-referencing, integration, aggregation or spatial
analysis (Burrough and McDonnell 1998). Many approaches have been developed to evaluate
aquifer vulnerability and can be grouped into three categories: (1) overlay and index methods;
(2) methods employing process-based simulation models, and (3) statistical methods
(National Research Council 1993, Tesoriero et al. 1998). Overlay and index methods combine
factors controlling the movement of pollutants from the ground surface into the saturated
zone resulting in vulnerability indices at different locations. Their main advantage is that
some of the factors such as rainfall and depth to groundwater can be available over large
areas, which makes them suitable for regional scale assessments (Thapinta and Hudak 2003).
However, their major drawback is the subjectivity in assigning numerical values to the
descriptive entities and relative weights for the different attributes. The process based
methods use simulation models to estimate the contaminant migration but they are
constrained by data shortage and computational difficulties (Barbash and Resek 1996; Rao
and Alley 1993).
Many methods for GVM, such as DRASTIC (Aller et al. 1987), GOD (Foster 1987), AVI
(Van Stempvoort et al. 1993), and SINTACS (Civita 1994), are able to distinguish degrees of
vulnerability at regional scales where different lithologies exist and have been mainly applied
to groundwater protection in porous aquifers.. A thorough overview of existing methods is
given in Vrba and Zaporozec (1994) and in Gogu and Dassargues (2000).
Since the conventional methods (i.e. DRASTIC, AVI, GOD, SINTACS) do not take into
account the peculiarities of karstic formations, other methods, such as, EPIK (Doerfliger and
Zwahlen 1998; Doerfliger et al. 1999), PI (Goldscheider et al. 2000) and COP (Vias et al.
2006), have been developed mainly for the assessment of vulnerability in carbonate (karstic)
aquifers.
Overlay and index methods and statistical methods are used to assess intrinsic
vulnerability, while methods employing process-based simulation models are used to assess
specific vulnerability. Models of index methods include GOD (Foster 1987), DRASTIC
(Aller et al. 1987), AVI rating system (Van Stempvoort et al. 1993), SEEPAGE, SINTACS,
ISIS (Gogu and Dassargues 2000), EPIK (Doerfliger et al. 1999), and DIVERSITY (Ray and
Odell 1993).
26 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

GVM for Porous Aquifers

DRASTIC Model

The most widespread method of evaluation of the intrinsic vulnerability is the DRASTIC
method (Aller et al. 1987). DRASTIC is an index model designed to produce vulnerability
scores for different locations by combining several thematic layers. It was originally
developed for manual overlay of semi quantitative data layers. However, the simple definition
of its vulnerability index as a linear combination of factors manifests the feasibility of the
computation using GIS (Fabbri and Napolitano 1995).
This method, taking into account seven parameters of the geological and hydrological
environment, was developed in the USA where it has been effectively applied several times
(Durnford et al. 1990; Evans and Myers 1990; Halliday and Wolfe 1991; Rundquist et al. 1991;
Fritch et al. 2000; Shukla et al. 2000), but also in many other regions of the world (Lobo-Ferreira
and Oliveira 1997; Lynch et al. 1997; Melloul and Collin 1998; Johansson et al. 1999; Kim and
Hamm 1999; Zabet 2002). The DRASTIC model was developed by the U.S. Environmental
Protection Agency (EPA) to evaluate groundwater pollution potential for the entire United States
(Aller et al. 1987). It was based on the concept of geological setting that is defined as a composite
description of all the major geologic and hydrologic factors that affect and control the groundwater
movement into, through and out of an area (Aller et al. 1987; Musa et al. 2000).

Table 1. Description of the DRASTIC model parameters (Babiker et al. 2005).

Relative
Factor Description Data Sources Data Type
Weight
Depth to the water table. Deeper water
Depth to water table levels imply lesser chance for 5 Borehole Data Point Data
contamination occurrences.
Water which penetrates the ground
surface and reaches the water table. Mean Annual
Recharge 4 Point Data
Recharge represents also the vehicle for Rainfall
transporting pollutants.
Properties of the saturated zone material
Geological
Aquifer media which control the pollutant attenuation 3 Polygon Data
Maps
processes.
The upper weathered portion of the
Soil media unsaturated zone which controls the 2 Soil Maps Polygon Data
amount of recharge that can infiltrate.
The slope of the land surface. It dictates Polyline Data
whether the runoff will remain on the Topographic (Contours) -
Topography 1
surface to allow contaminant Maps >Digital
percolation. Elevation Model
Unsaturated zone material. It controls
Impact of Geological/
the passage and attenuation of the 5 Polygon Data
vadose zone Soil Maps
contaminants.
The ability of the aquifer to transmit
Hydraulic Hydraulic
water and contaminants material within 3 Point Data
Conductivity Conductivity
the aquifer.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 27

The acronym DRASTIC stands for the seven parameters used in the model, which are:
Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone
media, and aquifer hydraulic Conductivity) model (Table 1).

Data Digitization

Thematic layers
Creation

Existing Maps

Assigning Rates for


different attributes

Remote Sensing Data Validation -


GIS Environment
Data Sensitivity Analysis
Assigning Weights
for different
thematic layers

Geophysical Data

Convert Vector
layers to Raster

Diffusion

Raster Calculations

Vulnerability Map

Figure 2. Flow chart of the main processing steps of GVM.

The model yields a numerical index that is derived from ratings and weights assigned to
the seven model parameters. The significant media types or classes of each parameter
represent the ranges, which are rated from 1 to 10 based on their relative effect on the aquifer
vulnerability. The seven parameters are then assigned weights ranging from 1 to 5 reflecting
their relative importance. The processing steps are shown in Figure (2).
The DRASTIC Index is then computed applying a linear combination of all factors
according to the following equation:

DRASTIC = Dr Dw + Rr Rw + Ar Aw + S r S w + Tr Tw + I r I w + C r C w

where D, R, A, S, T, I, and C are the seven aforementioned parameters and the subscripts r
and w are the corresponding rating and weights, respectively. DRASTIC uses a relatively
large number of parameters (seven parameters) to compute the vulnerability index which
28 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

ensures the best representation of the hydrogeological setting. The numerical ratings and
weights, which were established using the Delphi technique (Aller et al. 1987), are well
defined and are used worldwide. This makes the model suitable for producing comparable
vulnerability maps on a regional scale. The necessary information needed to build up the
several model parameters is, in general, available for the study area or it can easily be
inferred.
Several studies have used the DRASTIC model within a GIS environment although few
attempts have been made to apply the DRASTIC methodology in arid and semi-arid
environments (Fritch et al. 2000, GVM in Texas, USA).

Modifications of DRASTIC Model

Many modifications of DRASTIC model have been proposed by several authors, according to
data availability and problem singularity.
Evans and Myers (1990) used a GIS-based approach to evaluate the potential for regional
groundwater pollution with a modified DRASTIC approach in southern Delaware, USA.
Three DRASTIC parameters were not used in this research, namely net recharge, impact of
the vadose zone and the aquifer media. Instead, the authors added new parameters to the
DRASTIC index: the land use/land cover and septic tank system density. The authors claimed
that their approach could generate groundwater-related information for large geographical
areas that was sufficiently detailed for use by government agencies involved in protecting
groundwater.
Secunda et al. (1998) integrated the impact of extensive land use (risk) data over long
periods of time upon aquifer media as an additional parameter in the DRASTIC model, again
integrated into a GIS, to assess the potential level of groundwater vulnerability to pollution in
Israel’s Sharon region. The methodology employed empirical means to integrate aquifer
media and extensive agriculture land use data. Thus, the final assessment incorporated both
the natural state of the vadose zone and aquifer media (vulnerability) as well as the potential
danger posed by the long term effect upon the media of existing extensive land usage (risk) to
the region’s groundwater.
Piscopo (2001) used DRASTIC and GIS to produce a groundwater vulnerability map for
the Castlereagh Catchment in Australia. In this research, the author excluded hydraulic
conductivity from the final DRASTIC calculation due to the lack of data. Furthermore,
Piscopo (2001) replaced the recharge parameter (net recharge) as defined by the US EPA by
the potential of an area to have a recharge based on the rainfall amount, slope and soil
permeability.
Panagopoulos et al. (2006) proposed an optimization procedure of the original DRASTIC
method using various modifications and transformations on the basis of the statistical
parameters of a pollution index distribution. The pollution index which was used was the
nitrates concentration (expressed as mg/L NO3−) and the selection was based not only on the
fact that it constitutes the main contaminant that human activities introduce into the
environment of the study area, but also because it has been proposed as a representative
indicator of groundwater quality degradation (US EPA 1996). The DRASTIC parameters
were imported in a simple linear equation after they had been converted from the physical
range scale to a ten-grade relative scale. Each parameter is multiplied by a weighting
coefficient which had been determined with qualitative, not quantitative criteria, based on the
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 29

judgment of the authors of this method. The reduction of the physical range scale to the
relative ten-grade scale was conducted with the same philosophy. The linear equation of
determination had the following form:

Vint rinsic = D ⋅ λD + R ⋅ λR + A ⋅ λA + S ⋅ λS + T ⋅ λT + I ⋅ λI + C ⋅ λC

where Vintrinsic is the intrinsic vulnerability, D, R, A, S, T, I and C are the known DRASTIC
parameters and λ is the weighting coefficient for each factor. The major drawback of this
method is the subjectivity of the determination of the rating scale and the weighting
coefficients. Doubts have also been expressed for the selection of the specific parameters and
the exclusion of others. The detailed list of advantages/disadvantages of the DRASTIC
method can be found in Panagopoulos et al. (2006).

Application of the DRASTIC model to the Keritis Watershed of Crete Island

The DRASTIC model applied in GIS environment was used to evaluate the vulnerability of
the shallow porous aquifer in Keritis Basin. The major water use in Crete is for irrigation of
agriculture (84.5% of the total consumption), while domestic use is 12% and other uses 3.5%
(Chartzoulakis et al. 2001; Tsagarakis et al. 2004).
The study area is situated in the North Eastern part of Chania Municipality and
specifically from 35o24΄50΄΄N to 35o30΄00΄΄N, and 23o49΄50΄΄E to 23o58΄00΄΄E. The total
county area is 137 km2 and is located in the central part of Keritis river drainage basin, 3.5 km
west of the city of Chania. The central study area is characterized by a rather smooth
topography. The area is drained by Keritis river which is considered to be the main river of
the area.
The geological units were classified in the sense of permeability into four
hydrolithological units: high permeability rocks which comprise the karstic Triassic
limestones of Tripolis and Trypalion nappes, medium permeability rocks which consist of the
Quaternary deposits as well as the Miocene to Pliocene conglomerates and marly limestones,
low permeability rocks which consist of the Pliocene to Miocene marles and impervious
rocks which consist mainly of the phyllites – quartzites unit.
The local tectonic regime of the study area is characterized by faults of NW-SE and E-W
directions, which define the boundaries between the existing hydrolithological units as well as
the groundwater flow direction. Thus, these tectonic structures probably act as underground
dams bounding the underground water movement.
The average annual rainfall for the broader Chania area has been estimated to be 665 mm
(Chartzoulakis et al. 2001). About 65% of the annual precipitation is lost to
evapotranspiration, 21% as runoff to sea and only 14% recharges the groundwater
(Chartzoulakis et al. 2001). The rainfall is concentrated mainly in the winter months while the
drought period extends to more than six months (May to October). The monthly evaporation
ranges from 140 mm to more than 310 mm in the peak month. As a result, the water resources
availability is limited due to the spatio-temporal variations of precipitation (Tsagarakis et al.
2004). The demand for irrigation water is high, while at the same time only 31.0% of the
available agricultural land is irrigated (Tsagarakis et al. 2004). On the other hand, the
continuously growing tourist industry becomes more and more important in the local
30 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

economy and as a result the demand for high quality water becomes even stronger. The
growing water demands make the water resources modeling and management extremely
important for sustainable development.
In order to create the appropriate information platform upon which to proceed in a
systematic way towards applying the GVM, all available maps were collected (hydrological,
hydro-geological, geological, and topographic) and used as the basis for the creation of
several GIS thematic layers. Specifically, all the data were implemented into a GIS
environment and data digitization using ArcGIS software package was performed. The
several maps were geo-referenced to the local projection system of Greece (GGRS '87 -
Greek Geodetic Reference System) so that they could all be tied to the same projection
system, together with all future information that may become available.
The hydrological and hydro-geological data were collected during a Hydro-Geological
study of Chania prefecture and were provided by the Department of Hydrology of the
Ministry of Agriculture. All the available information concerning the wells, springs and
shafts of the broader area of Chania are gathered in these data, which are in the form of
maps and card inventories. The data contain important information such as the record code,
the geographical coordinates of the data, the municipality and the close-by cities or
villages, the maximum depth of the data (tubing depth, diameter of the tube, drilling depth,
depth of the groundwater level), the water supply, the usage, the geological stratigraphic
sketch of the data with a short lithological description and an in-situ time sequence of
measurements related to the depth of ground water, temperature and chemical
characteristics of the water (Cl- ions, EC- electrical conductivity). The locations of wells,
springs and shafts were digitized in the ArcGIS environment. Attribute tables (database)
have been created containing all the available hydro-geological data for each one of the
digitized locations.
The drainage networks were traced on transparency and digitized as hydrological maps
(1:20000) of the study area. The second principal component derived from the processing of a
Landsat–ETM satellite image of the study area, with a spatial resolution of 30x30 m pixel size
for the seven bands of multispectral data, acquired on 30 June 2000, was used for the
correction and update of the digitized stream data. The satellite data were also used for the
hydrolithological map production through the unsupervised classification with the Isodata
algorithm. The Digital Elevation Model (DEM) of the study area with a cell size of 20 m is a
continuous raster layer, in which data values represent elevation. DEM was generated from
the topographic maps of the study area.
The data that were finally used for the application of DRASTIC model are listed in
Table 2. After the creation of the primary layers, the polygon data were classified into the
certain classes (Table 2) and then they were converted to raster format (vector to raster
conversion). The point data (depth to groundwater table and hydraulic conductivity) were
interpolated using the Ordinary Kriging interpolator. Finally, the slope map was derived
from the DEM of the study area. The seven produced raster maps were reclassified using
the assigned rates. Finally, the reclassified layers were used as input parameters for the
raster calculator function.
In the application of DRASTIC on GIS environment, groundwater vulnerability is
estimated within raster/grid GIS. Raster models are cell-based representations of map
features, which offer analytical capabilities for continuous data and allow fast processing of
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 31

map layer overlay operations. In a raster GIS, the GV is calculated at a cell level as the
product of the seven factors

GVi = 5 Di ⋅ 4 Ri ⋅ 3 Ai ⋅ 2 S i ⋅ Ti ⋅ 5 I i ⋅ 3C i

where the subscript i represents the ith cell.

Table 2. Data origin, rates and weights used for the application of DRASTIC model in
the Keritis water basin.

Drastic Source Data Data form in Classes Rates Weights


Parameter GIS
0-1.5 10
1.5-4.5 9
4.5-9 7
Depth to Vector point
Card inventories 9-15 5 5
groundwater data
15-23 3
23-30.5 2
>30.5 1

High Permeability
9
Hydrogeological Medium
Vector 6
Recharge map/ Satellite Permeability 4
polygon data 3
imagery data Low Permeability
1
Impervious

Vector Alluvial 8
Aquifer type Geological Map 3
polygon data Limestone 10
Soil Geographical
Sandy Clay 4
Data Base of Europe Vector
Soil type Loam 5 2
(1:1,000,000) polygon data
Sandy Loam 6
(http://eusoils.jrc.it/)
>18 1
12-18 3
Topography Topographic maps
Raster data 12-6 5 1
(Slope (%)) (1:20000)
6-2 9
0-2 10

Phyllites-Quartzites 3
Impact of Geological Map Limestone 6
Polygon data 5
Vadoze Zone Neogene Sediments 6
Alluvial 8

0.01-1.3 1
1.3-3.9 2
Hydraulic 3.9-8.6 4
Geophysical Surveys Point data 3
Conductivity 8.6-13 6
13-24.2 8
>24.2 10
32 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Figure 3. Overlay of the seven raster maps representing the seven parameters of the DRASTIC model
and the raster calculator function which was applied.

Figure 4. The derived groundwater vulnerability map for the Keritis river basin. The major faults in the
area are shown with black lines.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 33

The groundwater vulnerability map (Figure 4) was the result of the operation shown in
Figure 3. As it is inferred from this map, the high hazard areas mainly governed by the depth
of the groundwater. Furthermore, the high hazard area includes the Koufos spring (upper left
part of the map), the medium quality of its water (is not potable), is confirmed by chemical
analysis and the Groundwater Quality Index application as presented in a following section
(Figure 8).

DRAMIC: A Modified Model of DRASTIC

DRASTIC method is probably the most widely used tool for vulnerability mapping, although
it is typically a poor predictor of groundwater contamination (Rupert 2001). It has been found
that the application of DRASTIC method in urban areas is limited.
First, the factor hydraulic conductivity is closely related to (or even more, is the most
important feature of) the factor aquifer media. Second, the topography of most cities is
relatively flat. Therefore, the factor topography can be ignored at urban areas in most cases.
Third, the ground surface of most of densely populated cities is covered by structures and
concrete, and it is difficult to obtain comparable values of the factor soil media for most
cities. To improve the predictability and applicability of the DRASTIC model, Wang et al.
(2007) proposed a new model: DRAMIC. The factors and assigned weights of DRAMIC in
comparison with those of DRASTIC are described in detail in Wang et al. (2007). The factor
topography is ignored. The factor soil media is substituted by a new factor, namely aquifer
thickness, and the factor hydraulic conductivity of the aquifer is substituted by a new factor,
namely impact of contaminant. The equation for determining the DRAMIC Index is:
DRAMIC

DRAMIC = 5 DR + 3 RR + 4 AR + 2 M R + 5 I R + 1C R

where R is rating. Once a DRAMIC Index has been calculated, it can be used to identify
areas, which are more likely to be susceptible to groundwater contamination relative to other.
The higher the DRAMIC Index is, the greater the groundwater pollution potential. The
hydrogeological significance, ranges and ratings for the factors D, R, A, and I are the same as
in DRASTIC. The ranges and ratings for the two new factors, namely thickness and
contaminant characteristics, are given in Wang et al. (2007). The water quantity in an aquifer
controls its ability to dilute contaminants and, therefore, affects the pollution potential.
Generally speaking, for the same region, the greater the aquifer’s thickness is, the lower the
vulnerability. As an intrinsic vulnerability assessment model, DRASTIC does not take into
account the effect of contaminant properties. Usually, however, the fate and transport of
different contaminants are different in the underground media. The type of contaminants, the
distance to the main aquifer, and the amount and pathway of contaminants all affect the
pollution potential, while the physico-chemical properties of contaminants determine their
stability or reactivity underground. Two major factors were considered in the DRAMIC
model: the contaminant’s stability and ease of infiltration into aquifer. Groundwater is usually
less vulnerable to unstable contaminants that cannot easily infiltrate into the aquifer.
34 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

SINTACS Method

The SINTACS method (Civita 1994; Civita and De Maio 2000; Corniello and Ducci 2000),
originally derived from DRASTIC, in the latest release 5, retains only the structure of
DRASTIC. It evaluates the vertical vulnerability using the same seven parameters: depth to
groundwater (S), recharge action (I), attenuation potential of the vadose zone (N), attenuation
potential of the soil (T), hydrogeologic characteristics of the aquifer (A), hydraulic
conductivity (C) and topographic slope (S). Each mapped factor is classified into ratings
(from 1 to 10) which have an impact on potential pollution. Weight multipliers are then used
for each factor to balance and enhance their importance. The final vulnerability index (Iv) is a
weighted sum of the seven factors. The weight classes used by SINTACS depend on the
hydrogeological features of each area. It is possible to use, in the same map, different weight
classes in different sectors.

2-STEP Method

The main objective of the method is the identification of maximum infiltration areas, taken as
groundwater infiltration zones, in order to delineate a protection strategy for municipal water
supplies (Brito et al. 2006). According to this method, the hydrogeological criteria that
describe the infiltration potential of present ground conditions in the study area should be
used.
Generally, the areas of maximum infiltration present the most favourable conditions for
recharge of aquifer systems. Considering that these areas are the most sensitive to
contamination risk, they are classified using criteria that reflect the vulnerability of the aquifer
system to possible contamination. Thus, the evaluation of maximum infiltration areas must
consider, directly or indirectly, the hydrogeological characteristics of the terrain, such as its
permeability and the hydrodynamics of the aquifer system. Other authors have used similar
methods to assess aquifer vulnerability (e.g. Aller et al., 1987; Francés et al., 2001). For the
application of the method, two steps must be followed: 1) characterization of local infiltration
potential throughout the area under study; 2) modeling and mapping the vulnerability of
groundwater to contamination. The final GV map results from the combination of maximum
infiltration map and the aquifer vulnerability indicator map. This final output provides
important information on the location of priority protection areas regarding the vulnerability
of the aquifer.
In the first step of application of the method, the hydrogeological parameters are
algebraically combined and the output map represents grades of high to low infiltration
potential. Four hydrogeological parameters are taken into account to locally characterize the
terrain infiltration potential: (i) geology, to describe substratum permeability, (ii) soil type, to
assess the permeability of non-consolidated sediments, (iii) slope, for the identification of
surface drainage areas, and (iv) water streams, to identify flooding zones. After the
delineation of infiltration potential map, the second step of the GIS model concerns the
discrimination of areas where the aquifer is most sensitive, taking into account the depth to
water table. At this step, Ordinary Kriging estimation methods should be used to improve
hydrostatic water level mapping and to evaluate the uncertainty of the estimation.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 35

GOD Method

This rating system method (Foster 1987) has a simple and pragmatic structure. It is an
empirical system for quick assessment of the aquifer vulnerability to pollution. According to
Gogu and Dassargues (2000), this method gives reliable results and is more suitable in
designing extended areas. Three main parameters are considered: the groundwater
occurrence, the lithology of the overlying layers, and the depth to groundwater (in unconfined
or confined conditions). The vulnerability index is the result of the values assigned to these
three parameters. Following the GOD flowchart, the vulnerability index of the area is
computed, by choosing first the rating of groundwater occurrence parameter and then
multiplying by the overlying lithology rating as well as with the depth to water parameter
rating. The overlying lithology parameter contributes to the vulnerability index only in the
case of unconfined aquifers.
Because the parameters can only take values from 0 to 1, the computation result is
usually a value less than the score assigned to each parameter. In the particular case where
two parameters have a value equal to 1, the vulnerability score is equal to the score of the
third parameter (Gogu and Dassargues 2000)

GVM for Karstic Aquifers

Groundwater from karst aquifers is among the most important resources of drinking water for
the growing population of the world. Carbonate rock outcrops, of which a large part is
karstified, cover about 7–12% of the planet’s dry, ice-free land, whereas karst waters supply
about 25% of the global population (Ford and Williams 1989). In Europe, carbonate terrains
occupy 35% of the land surface and a significant part of the drinking water supply is
abstracted from karstic/carbonate aquifers. In some European countries, karst water
contributes 50% to the total drinking water supply, and in many regions, it is the only
available fresh water resource. Recognizing these issues, the Directorate General for Science,
Research and Development of the European Commission supported COST Action 620
(COST is the acronym for “COoperation in Science and Technology) which considered
“Vulnerability and Risk Mapping for the Protection of Carbonate (Karst) Aquifers” which
was active between 1997 and 2003. This Action contributed to the development of the
European Water Framework Directive 2000/60/EC for river basin management and protection
of groundwater from pollution (DoELG/EPA/GSI 1999).
At the same time, karst aquifers are particularly vulnerable to contamination. Due to thin
soils, flow concentration in the epikarst (the uppermost, often intensively fractured and
karstified layer of a carbonate aquifer) and point recharge via swallow holes, contaminants
can easily reach the groundwater, where they may be transported rapidly in karst conduits
over large distances. The residence times of contaminants are often short, and processes of
contaminant attenuation often do not work effectively in karst systems (Goldscheider 2004).
Although the concept of groundwater vulnerability is applicable for all types of aquifers,
it was essential for the hydrologists to develop a method that takes into account the nature of
karst. Among the approaches they followed are the development of a method that is only
dedicated to karst and the development of a method that can be used for all types of aquifers,
but provides special tools for karst. There are three reasons why the above approaches are
36 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

more appropriate. First, there are transitions between purely fissured and karstified carbonate
aquifers. Secondly, there are transitions between porous and karst aquifers, e.g. intensively
fractured dolomites. Thirdly, there are often several types of interacting aquifers in one
hydrogeological system. The following characteristics of karst systems are relevant with
respect to groundwater vulnerability and, consequently, they should be taken into account
(compiled from Ford and Williams 1989; Drew and Hotzl 1999; Klimchouk et al. 2000;
Goldscheider 2004): 1) each karst system has its individual characteristics; generalizations are
doubtful; 2) karst systems are heterogeneous and anisotropic; thus, interpolation of data is
difficult and the reliability of a vulnerability map can be lower for karst than for other areas;
3) there are both diffusion and point recharge. Adjacent non-karst areas may generate surface
flow that may enter the karst aquifer via swallow holes (allogenic recharge); 4) the epikarst, if
present, controls the infiltration into the aquifer. It may store water and concentrate flow. The
structure and function of epikarst is often difficult to assess; 5) karst aquifers may comprise
conduits, fissures and intergranular pores. Contaminants can be transported very fast in the
conduits or stored in the fissures and pores (matrix); 6) karst systems show strong hydraulic
and physicochemical reactions to hydrologic events; 7) the water table and hydraulic gradient
are often difficult to define, particularly in shallow and conduit systems; 8) karst catchments
are often large and hydraulically connected over long distances. Karst catchments may
overlap and the flow paths (proved by tracer tests) may cross each other.
Within this context, EPIK (Doerfliger and Zwahlen 1998; Doerfliger et al. 1999) and PI
(Goldscheider et al. 2000) methods were specifically developed for the assessment of
vulnerability of carbonate/karstic areas.
The international approach to intrinsic karst groundwater vulnerability mapping considers
four factors. The overlying layers (O) may provide some degree of protection to the
groundwater. In karst areas, however, allogenic recharge may bypass these layers. Therefore,
the concentration of flow (C) also has to be considered. The precipitation regime (P) is
important when comparing groundwater vulnerability in different climatic regions but less
relevant for vulnerability mapping at a more local scale. The K factor describes the hydraulic
properties of the karst aquifer. Resource vulnerability maps are created by a combination of
the first three factors; source vulnerability maps additionally consider the K factor
(Goldscheider and Popescu 2004). Specific vulnerability maps can be created for different
types of contaminants, e.g. pesticides or pathogenic microorganisms (Sinreich et al. 2004).
Groundwater protection zones are areas in which land use is restricted in order to maintain
good water quality (Adams and Foster 1992). Some countries use vulnerability maps as a
basis for protection zoning. The importance of the groundwater is often considered as an
additional criterion (NRC 1993). In Ireland, for example, resource protection zones are
delineated on the basis of the vulnerability map and the importance of the aquifer, while
source protection zones are delineated on the basis of the vulnerability map and the travel
time in the aquifer.

The PI Method

The PI method is a GIS-based approach for mapping the intrinsic vulnerability of


groundwater resources (Goldscheider et al. 2000). It can be applied to all types of aquifers,
but provides special methodological tools for karst. The conceptual model of the method is
based on an origin-pathway-target model. The land surface is taken as the origin, the water
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 37

table in the aquifer is the target, and the pathway includes all layers in between. Vulnerability
is assessed as the product of two factors: protective cover (P) and infiltration conditions (I).
The detailed assessment schemes for the two factors can be found in Goldscheider et al.
(2000), Goldscheider (2004) and in the final report of the European COST Action 620
(Zwahlen 2004).
In general, the P factor describes the protective function of all layers that may be present
between the ground surface and the groundwater table: the topsoil (biologically most active
uppermost layer of the earth crust; pedologically the A and B soil horizons), the subsoil (non-
lithified sediments below the soil and over the bedrock; most often Quaternary deposits), the
non-karst rock and the unsaturated zone of the karst rock. Protectiveness is assessed on the
basis of the effective field capacity (eFC) of the soil, the grain size distribution (GSD) of the
subsoil, the lithology, fissuring and karstification of the non-karst and karst rock, the
thickness of all strata, the mean annual recharge and artesian pressure in the aquifer. The total
score range is divided into five classes, from P=1 for an extremely low degree of protection to
P=5 for very thick and protective overlying layers. A decadic logarithmic scale is applied, so
that a ten times higher protectiveness (e.g. 10-m-layer thickness instead of 1 m) makes the P
factor one class higher.
The I factor is new and crucial for the application of the method in karst areas. It
describes the infiltration conditions and, in particular, the degree to which the protective
cover is bypassed as a result of lateral surface and subsurface flow that enters the karst aquifer
at another place, for example via a swallow hole. The factor ranges between 0 and 1. It is 1 on
a horizontal, highly permeable soil, where all recharge will occur in a diffuse way, i.e. by
infiltration and subsequent percolation. In contrast, the I factor is 0 on a steep slope made of
low permeability soil that focuses surface runoff towards a sinking stream. In such a situation,
the protective cover will be completely bypassed. All other areas are assigned intermediate
values (0.2, 0.4, 0.6 and 0.8), depending on the soil properties controlling the predominant
flow process, the vegetation and slope gradient, and the position of a given point inside or
outside the catchment of a sinking stream.
The final protection factor p is the product of P and I. As proposed by Vrba and
Zaporozec (1994), five classes of vulnerability (or protectiveness) are distinguished: A
protective factor of p=1 indicates a very low degree of protection and an extreme
vulnerability to contamination, whereas a protective factor p=5 indicate a very high degree of
protection and a very low vulnerability.

The EPIK Method

After several tests with different methods, the EPIK method proved to be a suitable
parametric weight and point tool to quantify the vulnerability of different zones in jeopardy of
water contamination. Considering the karst aquifer’s geological, geomorphological and
hydrogeological characteristics (Doerfliger and Zwahlen 1995; Doerfliger et al. 1999), the
four parameters influencing flow and transport in karst taken into account by the method are
the following: epikarst (E), protective cover (P), infiltration condition (I) and karst network
development (K). Detailed information about the attribute classes for each of the parameters
is given in Barrocu et al. (2007). The four parameters categorized previously allow a
protection index value, F to be calculated for all parts of the catchments. The calculation is
carried out as follows,
38 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Fp i = αE i + bPi + cI i + dK i

i = 1,..,n is the grid cell number; Ei, Pi, Ii, Ki = weights assumed in the i-cell; a, b, c, d =
attribute relative weights (constant for any attribute); Fpi = i-cell protection factor (pertaining
to i-cell). The lower the protection factor calculated for any i-cell, the higher the vulnerability
of the karst aquifer. The steps for the EPIK method application are shown in Figure 5.

Land -use Topography Epikarst Geology Soils

DEM
Surface Runoff Geomorphological Aquifer Lithology
Coefficient &Vegetation Classification Classification
type

Slope
Classification

Overlay

Infiltration conditions Epikarst Rating Protective cover


Karst Rating (K)
Rating (I) (E) Rating (P)

Weighting Overlay
Vulnerability
Map

Map Sensitivity Analysis

Reclassification

Figure 5. Flowchart showing the successive steps for the EPIK method application.

The COP Method

The COP method has been developed for the assessment of intrinsic vulnerability of
carbonate aquifers. The acronym comes from the three initials of the factors used: flow
Concentration, Overlying layers and Precipitation. The conceptual basis of this method,
according to the European Approach (Daly et al. 2002; Goldscheider and Popescu 2004), is to
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 39

assess the natural protection of groundwater (O factor) determined by the properties of


overlying soils and the unsaturated zone, and also to estimate how this protection can be
modified by the infiltration process – diffuse or concentrated – (C factor) and the climatic
conditions (P factor – precipitation). In addition, the COP method establishes detailed
guidelines, tables and formulae for vulnerability assessment and selects the variables,
parameters and factors to be used according to the European Approach proposed by COST
Action 620 (Daly et al. 2002; Zwahlen 2004). The method can be applied using geo-
environmental data available in most countries, with some fieldwork but without extensive
input from geographical information systems (GIS). As a result the COP method is likely to
be practical and useful for decision makers implementing groundwater protection schemes.
The method can be applied in different climatic conditions and different types of
carbonate aquifers (diffuse and conduit flow systems). In addition, the COP method uses
variables, parameters and factors in line with those proposed in the European Approach (Daly
et al. 2002; Zwahlen 2004) and it can be applied using different levels of available data.
The factors of the COP method have been combined to evaluate the intrinsic vulnerability
of a groundwater resource, as proposed in the following formula:

COP Index = C · O · P

The final numerical representations of the C, O and P factors (the C, O and P scores) are
multiplied, because each one is considered to impact on the assessment of vulnerability of
karst aquifers. Within the COP method, the values for the intrinsic vulnerability index range
between 0 and 15. Following the proposal by Vrba and Zaporozec (1994), the values for this
index are grouped into five vulnerability classes (Very High, High, Moderate, Low and Very
Low vulnerability).
Vias et al. (2006) applied in two aquifers in the Southern Spain, all the most frequent
GVM methods such as, DRASTIC, GOD, AVI, SINTACS (Longo et al. 2001), EPIK, COP
and PI (Brechenmacher 2002; Vıas et al. 2005; Andreo et al. 2006). Each of these studies
concluded that the COP method was the most effective for assessing the prevailing
vulnerability of these aquifers based on actual hydrogeological understanding of the aquifers.

Qualitative Characterization of Aquifers

The task of assessing the quality of an aquifer as an entirety and to conclude about its
vulnerability to contamination becomes very problematic due to difficulties originating from
spatial and temporal variability of contamination sources, inhomogeneity of solid material
above the aquifer, lack of knowledge regarding attenuation processes of the pollutants, wrong
estimation of recharge, etc. Additionally, besides the abovementioned difficulties, monitoring
the overall quality of water (surface or groundwater) includes a large number of potential
pollutants. Specific determination of each of them could result in very high cost and time
consumption. Alternatively, a number of suitable water quality indicators could be chosen and
monitored, which finally will be used to develop a groundwater quality index.
40 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

The Groundwater Quality Index, GQI

Based on the abovementioned reasons Babiker et al. (2007), developed a GIS-based


groundwater quality index (GQI) combining different available water quality data (e.g., Cl−,
Na+, Ca2+ concentration) and linked them to the World Health Organization (WHO)
standards. They applied the proposed index to Nasuno basin, Japan and they related the final
GQI map with more data such as the depth to groundwater table, the geomorphologic
structures, the land-use types and population density in order to find out which factors control
the spatial variability of groundwater quality.
The main part of the GQI represents an averaged linear combination of factors. The
weight (w) assigned to each parameter indicating its relative importance to groundwater
quality, corresponds to the mean rating value of its “rank map”. Parameters which have
higher impact over groundwater quality (high mean rate) are assumed to be similarly more
important in evaluating the overall groundwater quality. Particular emphasis was given to
contaminants that posses potential risk to human health (w = mean r + 2). High GQI values
close to 100 reflect high water quality and index values far below 100 (close to 1) indicate
low water quality (Babiker et al. 2007).
A similar approach, the IAWQ (Indexed Assessment of an aquifer’s Water Quality) index
was also suggested by Melloul and Collin (1998), however their basic assumptions and
formulations are generally different from those introduced by Babiker et al. (2007).

Monitoring of Water Quality of Keritis River Basin

The sampling sites chosen for this study are marked on the map of Figure 8. More
specifically, samples were taken from four springs in the area of Agyia, (altitude ~40 m above
the sea level), as well as from three springs in the area of Meskla village (altitude ~200 – 220
m), Fassa's spring and the drillings of Koufos, Myloniana and Fournes. The samples
collected from springs out of the Agyia zone were analyzed for two reasons: first because the
waters from these sites are also used as supplies for the population and the agriculture of the
area and secondly, in order to compare the quality of their water with the quality of the water
of Agyia springs.
Water samples were collected on a monthly basis from August 2005 until April 2007.
Sampling and sample preservation were done according to methods 1060 B and 1060 C of
"Standard Methods for the Examination of Water and Wastewater" (APHA 2005). Samples
were subjected to several water quality analyses and the results were stored in a data base.
The groundwater quality index was generated using seven parameters (Cl−, Na+, Ca2+,
Mg2+, NO3-, SO42− , and total dissolved solids, TDS), which are listed in World Health
Organization (WHO) and European Community (EC) guidelines for drinking water quality
(WHO 2004; EC 1998). Except nitrate ions, which have been accused as a "potential health
risk" chemical and an upper concentration limit of 50 mg/L has been set, all the other
parameters are not considered to be dangerous for human health (the possibility of laxative
effects of water with very high concentrations of sulfates combined with magnesium or
sodium should probably be mentioned at this point) (WHO 2004; EC 1998). For this reason,
WHO and EC have not established fixed limits for these chemicals, but instead they provide
specific guideline levels for them (WHO 2004) or they consider them as water quality
indicator parameters with values set for monitoring purposes (EC 1998). However, their
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 41

existence in water influences its taste and odor, and consequently, its acceptability by
consumers.
The chemical parameters used for the development of the groundwater quality index may
originate either from natural sources and/or from anthropogenic activity, i.e. Ca2+ and Mg2+
originate mainly from the weathering of several mineral deposits like chalk, limestone or
dolomites etc; SO42- can come from dissolution of gypsum and other mineral deposits
containing this ion; Na+ occurs with silicates and other salt deposits; Cl- (as well as Na+) can
be found in groundwaters which have been affected by seawater intrusion or by agricultural
or irrigation discharges; NO3- may come from the extensive use of nitrogenous fertilizers.
Regardless their origin, these and other chemicals may be transferred through the soil, the
unsaturated and saturated zone towards aquifers and affect the quality of groundwater. This
process is, in general, very complicated and it is governed mainly by the physico-chemical
properties of the soil and parent rock, as well as by the properties of the pollutant and its
potential interactions with soil/rock which may result in attenuation of the pollutants.

Application of GQI in Keritis River Basin

The GQI method was applied for the Keritis basin area which encompasses the Agyia lake
(Figure 6), which is of great importance mainly due to the limited number of freshwater
bodies in Crete, the important role of freshwater biotopes for the water balance and the
biodiversity of the island of Crete and the importance of the wetland as a refugee for many
bird species.
At a first step, seven concentration maps representing the “primary map I” were
constructed for each chemical parameter from the point data using the Ordinary Kriging
interpolation method. In a next processing phase, and in order to relate the data to universal
norm, the measured concentration, X_, of every pixel in the “primary map I” was related to
its desired WHO standard value, X (Table 3), using a normalized difference index:

C=
(X _− X )
X _+ X

Table 3. Statistics of the seven parameters of groundwater quality from the Keritis
basin and the corresponding maximum thresholds guidelines according to the WHO
and to EC. All parameters are given in mg/L.

Min Max Mean SD WHO EC


2+
Ca 0.70 174.70 43.64 56.72 300 -
Mg2+ 0.00 36.14 7.78 10.06 300 -
Na+ 3.15 29.40 10.27 7.24 200 200
Cl– 16.00 48.00 25.15 10.76 250 250
NO3– 0.60 3.20 1.52 0.81 50 50
SO42– 25.00 1500.00 279.09 512.86 250 250
TDS 16.40 918.40 263.12 296.86 600 -
42 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Figure 6. 3,5,7 (R,G,B) Landsat-ETM color composite superimposed to the Digital Elevation Model of
the Keritis basin. Overlaid linear vector layer of the major lineaments is also shown with red color.

The seven resultant “primary map II” display for each pixel contamination index values
ranging between −1 and 1. This is close to the contamination index approach which is
calculated as the ratio between the measured concentration of a contaminant and the
prescribed maximum acceptable contaminant level (Melloul and Collin 1998; Praharaj et al.
2002; Babiker et al. 2005). However, the normalized difference index proposed by Babiker et
al. (2007) provides fixed upper and lower limits for the contamination level.
In a third step, the contamination index (primary map II) was rated between 1 and 10 to
generate the “rank map” using the following polynomial function which ranks the
contamination level (C) of every pixel between 1 and 10:

r = 0.5 ⋅ C 2 + 4.5 ⋅ C + 5

where, C represents the contamination index value for each pixel and r represents the
corresponding rank value (Figure 7). After that, the relative weight (w) of each parameter has
to be calculated as the mean value (r: 1–10) of the corresponding rank map and the “mean r +
2” (r ≤ 8) for parameters that have potential health effects (e.g. nitrate) (Table 4).
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 43

In a final stage, and using the raster calculator functions, the GQI map (Figure 8)
concerning the seven chemical parameters was calculated as follows:

⎛ (r w + r w + ... + r7 w 7 ) ⎞
GQI = 100 − ⎜ 1 1 2 2 ⎟
⎝ 7 ⎠

where, r represents the rate of the rank map and w represents the relative weight of the
parameter.
The resulted map, gave groundwater quality indices between 92.28 and 96.94. These
values were classified into two equal classes. Colors were then assigned to each one of the
classes (Figure 8). The green color indicates “Maximum” water quality, while the yellow
color indicates “Medium” water quality. This representation was chosen because it clearly
demonstrates the significant role of the existed major fault which is considered to act as a
barrier for the two groundwater qualities. This finding becomes even more impressive if one
considers that the distance between the two neighboring springs with different water quality is
a few hundred meters.
It should be noted that water from both areas as they are depicted by green and yellow
colours in Figure 8 meet most of the guideline values for the corresponding parameters set by
WHO. However, there is a clear distinction between the water sampled from the "yellow"
sampling sites and the water sampled from the "green" sampling sites regarding the
concentrations of Ca2+, Mg2+, SO42-, Cl-, as well as the amount of TDS, with the former
showing much higher concentrations than the latter. Especially for SO42-, water from the
"yellow" sampling sites showed concentrations well above the WHO and EC guidelines.
Based on these facts one can consider this water as of lesser quality compared to the other.

Figure 7. Overlay of the seven rank maps. Rate values equal or near to 1 are depicted with blue colors
and indicate minimum impact on groundwater quality, while medium and maximum impact on
groundwater quality is shown with warm (red) colors.
44 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 4. Summary of the statistics of the seven rank maps used to generate the
groundwater quality index in Keritis basin. The mean rank values were used as
weighting factors. Note that for nitrate ions the weighting factor was mean+2. In the last
row, the statistics for the final GQI map are given.

Parameter Minimum Maximum Mean SD


Ca2+ 1.012 3.85 1.5 0.52
Mg2+ 1 1.78 1.15 0.12
Na+ 1.11 1.93 1.3 0.105
Cl– 1.53 2.43 1.81 0.16
NO3– 1.08 1.43 1.22+2 0.067
SO42– 1.65 8.47 2.64 1.39
TDS 1.19 5.67 2.41 0.88
GQI 92.28 96.94 96.38 0.99

Figure 8. The groundwater quality index map (GQI) of the Keritis basin, superimposed to the Digital
Elevation Model of the Keritis basin. The continuous red solid line indicates the visible fracture zone
and the dashed line the possible continuation of the tectonic line. The resulted "quality barrier"
coincides with the fault shown with the red line.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 45

Groundwater Flow Models and GIS

The use of groundwater models is prevalent in the field of environmental science. Models
have been applied to investigate a wide variety of hydrogeologic conditions. More recently,
groundwater models are being applied to predict the transport of contaminants for risk
evaluation.
In general, models are conceptual descriptions or approximations that describe physical
systems using mathematical equations; they are not exact descriptions of physical systems or
processes. By mathematically representing a simplified version of a hydrogeological system,
reasonable alternative scenarios can be predicted, tested, and compared. The applicability or
usefulness of a model depends on how closely the mathematical equations approximate the
physical system being modeled. In order to evaluate the applicability or usefulness of a
model, it is necessary to have a thorough understanding of the physical system and the
assumptions embedded in the derivation of the mathematical equations.
Groundwater models describe the groundwater flow and transport processes using
mathematical equations based on certain simplifying assumptions (Torak 1992a,b; Cooley
1992; Lin Y-F et al. 2003). These assumptions typically involve the direction of flow,
geometry of the aquifer, the heterogeneity or anisotropy of sediments or bedrock within the
aquifer, the contaminant transport mechanisms and chemical reactions. Because of the
simplifying assumptions embedded in the mathematical equations and the many uncertainties
in the values of data required by the model, a model must be viewed as an approximation and
not an exact reconstruction of the real field conditions. Groundwater models, however, even
as approximations are a useful investigation tool that groundwater hydrologists may use for a
number of applications.
Nowadays, water managers and policy makers require studies that organize the data,
evaluate the system, improve understanding, and establish a framework for follow-up
analyses. This effort is assisted by the use of GIS and database technologies, which provide
efficient tools for building a knowledge base for this study and translating geohydrological
data into model-ready formats. Examples are provided for several procedures to efficiently
manage, interpolate, and reformat MODFLOW data using ArcGIS software (Goodchild et al.
1993; Srinivasan and Arnold 1994; Price and Pierce 1994; Watkins et al. 1996; Orzol 1997;
Lasserre et al. 1999; Gogu et al. 2001; Gossel et al. 2004; Chatterjee et al. 2005).
Modeling groundwater flow and contaminant transport in aquifers represents a spatial
and temporal problem that requires the integration of deterministic process-based models with
GIS. In order to model the physical and chemical processes in the aquifer, each model
parameter or variable of the model is represented on a three or four dimensional (x, y, z, and
time) information layer. Due to the heterogeneity of aquifers, representing the spatial
distribution of the parameters and variables that are involved in the constitutive laws
describing the simulated processes creates a huge data volume. Managing these data can be
done most effectively through GIS. Moreover, GIS simplify significantly the implementation
of data management tasks of model building and model calibration.
In this section, in an effort to aid geologists/hydrogeologists to make their work easier, a
GIS integrated groundwater flow and contaminant fate and transport modeling platform is
presented using ArcGIS and the GroundWaterVistas software. The GIS platform facilitates
the time consuming task of preparation of data input and output structures for multilayer
groundwater flow and contaminant fate and transport simulation codes.
46 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Basics of Groundwater Flow and Mass Transport

Groundwater modeling begins with a conceptual understanding of the physical problem. The
next step in modeling is translating the physical system into mathematical terms (Toth 1963;
Freeze and Witherspoon 1966, 1967; Bredehoeft and Pinder 1973; Bennett 1976; Anderson
1984; Cherry et al. 1984; Franke et al. 1991; Knox et al. 1993; Hanson and Leake 1999;
Merritt and Konikow 2000). In general, the results are the familiar groundwater flow equation
and transport equations. The governing flow equation for three-dimensional saturated flow in
saturated porous media is:

∂ ⎛ ∂h ⎞ ∂ ⎛ ∂h ⎞ ∂ ⎛ ∂h ⎞ ∂h
⎜ K xx ⎟ + ⎜⎜ K yy ⎟⎟ + ⎜ K zz ⎟ − Q = SS
∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠ ∂z ⎝ ∂z ⎠ ∂t

where, Kxx, Kyy, Kzz are the hydraulic conductivities along the x,y,z axes which are assumed
to be parallel to the major axes of hydraulic conductivity, h is the piezometric head, Q is the
volumetric flux per unit volume representing source/sink terms and Ss is the specific storage
coefficient defined as the volume of water released from storage per unit change in head per
unit volume of porous material.
The transport of solutes in the saturated zone is governed by the advection dispersion
equation which, for a porous medium with uniform porosity distribution, is formulated as
follows:

∂c ∂ ⎛ ⎞
=− (cv i ) + ∂ ⎜⎜ Dij ∂c ⎟ + Rc , i,j=1,2,3

∂t ∂x i ∂x i ⎝ ∂x j ⎠

where, c is the concentration of the solute, Rc is the sources or sinks, Dij is the dispersion
coefficient tensor and vi is the velocity tensor.
An understanding of these equations and their associated boundary and initial conditions
is necessary before a modeling problem can be formulated. Basic processes that are
considered include groundwater flow, solute transport and heat transport. Most groundwater
modeling studies are conducted using either deterministic models, based on precise
description of cause-and-effect or input-response relationships, or stochastic models reflecting
the probabilistic nature of a groundwater system.
The governing equations for groundwater systems are usually solved either analytically
or numerically. Analytical models contain analytical solution of the field equations,
continuously in space and time. In numerical models, a discrete solution is obtained in both
the space and time domains by using numerical approximations of the governing partial
differential equation. Various numerical solution techniques are used in groundwater models.
Among the most used approaches in groundwater modeling, three techniques can be
distinguished: Finite Difference Method, Finite Element Method, and Analytical Element
Method. All techniques have their own advantages and disadvantages with respect to
availability, costs, user friendliness, applicability, and user background.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 47

Groundwater Model Application

A groundwater model application is the combination of two different processes (Figure 9).
The first process is the model development resulting in a software product, and the second
process is the application of that product for a specific purpose. Groundwater models are most
efficiently developed in a logical sequence. In order to build a complete and well organized
model, specific model steps should be followed as described below (Pinder and Bredehoeft
1968; Wang and Anderson 1982; Kinzelbach 1986; Bear and Verruijt 1987; McDonald and
Harbaugh 1988; Franke et al. 1991; Anderson and Woessner 1992; ASTM 5447-93; ASTM
1689-95; ASTM 5880-95; Ghosh and Sharma 2006):

Figure 9. Typical flow chart of the model development (left flowchart) and model application (right
flowchart).

Model Objectives (MO). Model objectives which explain the purpose of using a groundwater
model should be defined (Harbaugh et al. 2000).

Hydrogeological Characterization (HC). In order to select an appropriate model or develop a


reliably calibrated model, a detailed characterization of the hydrogeological conditions in the
study area is needed, so the importance of relevant flow or solute transport processes will be
understood.

Model Conceptualization (MC). Model conceptualization is the process in which data


describing field conditions are assembled in a systematic way to describe groundwater flow
and contaminant transport processes. The MC helps in choosing the proper modeling
approach and the model software to use.
48 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Modeling Software Selection (MSS). After HC of the site has been completed and the MC is
developed, a computer model software is should be selected. This model should be capable to
simulate conditions encountered at a site.

Model Design (MD - Input Parameters). MD includes all parameters that are used to develop
a calibrated model. The input parameters include model grid size and spacing, layer
elevations, boundary conditions, hydraulic conductivity/transmissivity, recharge, any
additional model input, transient or steady state modelling, dispersion coefficients,
degradation rate coefficients, etc (Franke et al. 1987; ASTM 5610-94).

Model Calibration (MoC). MoC consists of changing values of model input parameters in an
attempt to match field conditions within some acceptable criteria. Model calibration requires
that field conditions at a site are properly characterized (Konikow 1978; Fryberg 1988;
ASTM 5490-93; Bredehoeft and Konikow 1993; ASTM 5918-96; Hill 1998).

Sensitivity Analysis (SA). A SA is the process of varying model input parameters over a
reasonable range (range of uncertainty in value of model parameter) and observing the
relative change in model response (ASTM 5611-94; Hill et al. 2000). Data for which the
model is relatively sensitive should require further characterization.

Model Verification (MV). A calibrated model uses selected values of hydrogeologic


parameters, sources and sinks and boundary conditions to match historical field conditions.
After the MV step, the model is ready for predictive simulations.

Predictive Simulations (PS). A model may be used to predict some future groundwater flow
or contaminant transport condition (Gleeson 1967; Fryberg 1988). The model may also be
used to evaluate different remediation alternatives.

Performance Monitoring Plan (PMP). Groundwater models are used to predict the migration
pathway and concentrations of contaminants in groundwater. Small errors in the PS can be
propagated in solutions projected forwarded in time.

Groundwater Flow Model and Data Used

A catalog of 50 most frequently used groundwater flow models (GWFM) and the overall
characteristics/features advantages/disadvantages of the GWFM is given below (Table 5).
The most widely used numerical groundwater flow model is MODFLOW which is a three-
dimensional model, originally developed by the U.S. Geological Survey (McDonald and
Harbaugh 1988). It uses block-centered finite difference scheme for saturated zone. The
advantages of MODFLOW include numerous facilities for data preparation, easy exchange of
data in standard form, extended worldwide experience, continuous development, availability
of source code, and relatively low price. However, surface runoff and unsaturated flow are
not included, hence in case of transient problems MODFLOW can not be applied if the flux at
the groundwater table depends on the calculated head and the function is not known in
advance.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 49

Table 5. Short description of the most frequently used GWFM


(Ghosh and Sharma 2006)

GWFM Description
3DFEMFAT is a 3-Dimensional Finite-Element Model of Flow And Transport
through Saturated-Unsaturated Media. Typical applications are infiltration,
wellhead protection, agriculture pesticides, sanitary landfill, radionuclide
disposal sites, hazardous waste disposal sites, density-induced flow and transport,
saltwater intrusion, etc. 3DFEMFAT can do simulations of flow only, transport
only, combined sequential flow and transport, or coupled density-dependent flow
and transport. In comparison to conventional finite-element or finite-difference
models, the transport module of 3DFEMFAT offers several advantages: (1) it
3DFEMFAT
completely eliminates numerical oscillation due to advection terms, (2) it can be
applied to mesh Peclet numbers ranging from 0 to infinity, (3) it can use a very
large time step size to greatly reduce numerical diffusion, and (4) the hybrid
Lagrangian-Eulerian finite-element approach is always superior to and will never
be worse than its corresponding upstream finite-element or finite-difference
method. Because of these advantages, 3DFEMFAT is suitable for applications to
large field problems. It is flexibile and versatile in modeling a wide range of real
world problems.
AQUA3D is a program developed to solve three-dimensional groundwater flow
and transport problems using the Galerkin finite-element method. AQUA3D
solves transient groundwater flow with inhomogeneous and anisotropic flow
conditions. Boundary conditions may be prescribed nodal head and prescribed
AQUA3D
flow as a function of time or head-dependent flow. AQUA3D also solves
transient transport of contaminants and heat with convection, decay, adsorption
and velocity-dependent dispersion. Boundary conditions may be either prescribed
nodal concentration (temperature) or prescribed dispersive mass (heat) flux.
Argus ONE is an advanced graphical preprocessing and post-processing software
that seamlessly integrates with ground-water models, GIS and work flow. Argus
ONE gives you a unified modeling solution for all your ground-water problems.
The Argus unique Plug-In Extension (PIE) technology enables anyone to customize
Argus ONE Argus ONE as a graphical user interface for their models. Using this technology,
many organizations, including the USGS, created customized Argus ONE Plug-Ins
as GUIs for their ground-water models. Interfaces for the following models are
available: MODFLOW, MOC3D, MODPATH, HST3D, SUTRA, PTC,
ARC/INFO, ArcView
ChemFlux and Chemflo is a stable finite element contaminant transport modeling
software. It is a finite element software package characterized by automatic mesh
generation, automatic mesh refinement and automatic time-step refinement. The
solver offers speed and reduction in convergence problems. Results of
benchmark tests, run against MT3D, confirm the effectiveness of the solver.
ChemFlux is able to provide the same level of accuracy as MT3D in solutions
ChemFlux dominated by advection while implementing the irregular geometry benefits of
the finite element method. ChemFlux can also import groundwater gradients
from the SVFlux groundwater modeling package. Prediction of the movement of
contaminant plumes through the processes of advection, diffusion, adsorption
and decay is possible. The ChemFlux design module provides an elegant and
simple user interface. Problem geometry and groundwater gradients may be
imported from the SVFlux software.
50 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 5. Continued

GWFM Description
FEFLOW is a finite-element package for simulating 3D and 2D fluid density
coupled flow, contaminant mass (salinity) and heat transport in the
subsurface. It is capable of computing: 1) Groundwater systems with and
without free surfaces (phreatic aquifers, perched water tables, moving
meshes), 2) Problems in saturated-unsaturated zones, 3) Both salinity-
FEFLOW dependent and temperature-dependent transport phenomena (thermohaline
flows), 4) Complex geometric and parametric situations. The package is fully
graphics-based and interactive. Pre-, main- and post-processing are
integrated. There is a data interface to GIS and a programming interface. The
implemented numerical features allow the solution of large problems.
Adaptive techniques are incorporated.

FLONET/TRANS is a software package for 2-D cross-sectional groundwater


flow and contaminant transport modeling. The modeling environment offers
all the advantages of finite-element modeling (numerical stability and
flexible geometry) together with a logical and intuitive graphical interface
that makes finite-element modeling fast and easy. It uses the dual formulation
FLONET/TRANS
of hydraulic potentials and streamlines to solve the saturated groundwater
flow equation and creates accurate flownet diagrams for any two-
dimensional, saturated groundwater flow system. In addition, it simulates
advective-dispersive contaminant transport problems with spatially-variable
retardation and multiple source terms.

FLOWPATH for Windows is a popular model for groundwater flow,


remediation, and wellhead protection. It is a comprehensive modeling
environment, specifically designed for simulating 2-D groundwater flow and
contaminant transport in unconfined, confined and leaky aquifers with
heterogeneous properties, multiple pumping wells and complex boundary
FLOWPATH
conditions. Some typical applications of FLOWPATH include: 1)
Determining remediation well capture zones, 2) Delineating wellhead
protection areas, 3) Designing and optimizing pumping well locations for
dewatering projects and 4) Determining contaminant fate and exposure
pathways for risk assessment.

GFLOW is an efficient stepwise groundwater flow modeling system. It is a


Windows 95/98/NT program based on the analytic element method. It
models steady-state flow in a single heterogeneous aquifer using the Dupuit-
Forchheimer assumption. While GFLOW supports some local transient and
three-dimensional flow modeling, it is particularly suitable for modeling
GFLOW regional horizontal flow. To facilitate detailed local flow modeling, it
supports a MODFLOW-extract option to automatically generate
MODFLOW files in a user-defined area with aquifer properties and boundary
conditions provided by the GFLOW analytic element model. GFLOW also
supports conjunctive surface water and groundwater modeling, using stream
networks with calculated baseflow.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 51

Table 5. Continued

GWFM Description
GMS is a sophisticated and comprehensive groundwater modeling software. It
provides tools for every phase of a groundwater simulation including site
characterization, model development, calibration, post-processing, and
visualization. GMS supports both finite-difference and finite-element models
GMS in 2D and 3D including MODFLOW 2000, MODPATH, MT3DMS/RT3D,
SEAM3D, ART3D, UTCHEM, FEMWATER and SEEP2D. The program's
modular design enables the user to select modules in custom combinations,
allowing the user to choose only those groundwater modeling capabilities that
are required.
Groundwater Vistas (GV) is a sophisticated windows graphical user interface for
3-D groundwater flow and transport modeling. It couples a model design system
with comprehensive graphical analysis tools. GV is a model-independent
graphical design system for MODFLOW MODPATH (both steady-state and
transient versions), MT3DMS, MODFLOWT, MODFLOWSURFACT,
MODFLOW2000, GFLOW, RT3D, PATH3D, SEAWAT and PEST, the model-
independent calibration software. The combination of PEST and GV's automatic
sensitivity analysis make GV a good calibration tool. The advanced version of
Groundwater
Groundwater Vistas provides the ideal groundwater risk assessment tool.
Vistas
Groundwater Vistas is a modeling environment for the MODFLOW family of
models that allows for the quantification of uncertainty. Stochastic (Advanced)
Groundwater Vistas includes, Monte Carlo versions of MODFLOW,
MODPATH and MT3D, Geostatistical Simulators SWIFT support advanced
output options and more. GV displays the model design in both plan and cross-
sectional views using a split window (both views are visible at the same time).
Model results are presented using contours, shaded contours, velocity vectors,
and detailed analysis of mass balance.
The Windows version of MicroFEM is a new software package based on the
DOS version Micro-Fem. It takes you through the whole process of ground-
water modeling, from the generation of a finite-element grid through the stages
of preprocessing, calculation, post-processing, graphical interpretation and
plotting. Confined, semi-confined, phreatic, stratified and leaky multi-aquifer
systems can be simulated with a maximum of 20 aquifers. Irregular grids, as
typically used by finite-element programs, have several advantages compared
to the more or less regular grids used by finite-difference codes. A model with
a well-designed irregular grid will show more accurate results with fewer
nodes, so less computer memory is required, while calculations are faster.
MicroFEM MicroFEM offers extensive possibilities as to the ease of creating such
irregular grids. Other MicroFEM features include the ease of data preparation
and the presentation and analysis of modeling results. A flexible way of zone-
selection and formula-assignment is used for all parameters: transmissivities,
aquitard resistances, well discharges and boundary conditions for each layer.
Depending on the type of model, this can be extended with layer thicknesses,
storativities, spatially varying anisotropy, topsystem and user-defined
parameters. To inspect and interpret model results, maps and profiles can be
used to visualize contours, heads, 3D-flowlines, flow vectors, etc. Time-
drawdown curves and water balances can be selected with just a few
keystrokes or mouse clicks.
52 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 5. Continued

GWFM Description
MOC simulates 2D/3D solute transport in flowing groundwater. MOC is both
general and flexible in that it can be applied to a wide range of problem types.
MOC is applicable for one- or two-dimensional problems involving steady-state
or transient flow. MOC computes changes in concentration over time caused by
the processes of convective transport, hydrodynamic dispersion, and mixing (or
dilution) from fluid sources. MOC assumes that gradients of fluid density,
viscosity and temperature do not affect the velocity distribution. However, the
MOC aquifer may be heterogeneous and/or anisotropic. MOC is based on a
rectangular, block-centered, finite-difference grid. It allows the specification of
injection or withdrawal wells and of spatially-varying diffuse recharge or
discharge, saturated thickness, transmissivity, boundary conditions and initial
heads and concentrations. MOC incorporates first-order irreversible rate-
reaction; reversible equilibrium controlled sorption with linear, Freundlich, or
Langmuir isotherms; and reversible equilibrium-controlled ion exchange for
monovalent or divalent ions.
MOCDENSE is a modified version of the ground-water flow and solute-transport
model of Konikow and Bredehoeft which was designed to simulate the transport
and dispersion of a single solute that does not affect the fluid density. This
modified version of MOCDENSE simulates the flow in a cross-sectional plane
rather than in an areal plane. Because the problem of interest involves variable
MOCDENSE
density, the modified model solves for fluid pressure rather than hydraulic head in
the flow equation; the solution to the flow equation is still obtained using a finite
difference method. Solute transport is simulated in MOCDENSE with the method
of characteristics as in the original model. Density is considered to be a function
of the concentration of one of the constituents.
MODFLOW is the name that has been given the USGS Modular Three-
Dimensional Ground-Water Flow Model. Because of its ability to simulate a wide
variety of systems, its extensive publicly available documentation and its rigorous
USGS peer review, MODFLOW has become the worldwide standard GWFM.
MODFLOW is used to simulate systems for water supply, containment
remediation and mine dewatering. When properly applied, MODFLOW is the
recognized standard model. The modular structure of MODFLOW consists of a
Main Program and a series of highly-independent subroutines called modules. The
modules are grouped in packages. Each package deals with a specific feature of
the hydrologic system which is to be simulated such as flow from rivers or flow
into drains or with a specific method of solving linear equations which describe
MODFLOW
the flow system such as the Strongly Implicit Procedure or Preconditioned
Conjugate Gradient. The division of MODFLOW into modules permits the user to
examine specific hydrologic features of the model independently. This also
facilitates development of additional capabilities, because new modules or
packages can be added to the program without modifying the existing ones. The
input/output system of MODFLOW was designed for optimal flexibility. Ground-
water flow within the aquifer is simulated in MODFLOW using a block-centered
finite-difference approach. Layers can be simulated as confined, unconfined, or a
combination of both. Flows from external stresses such as flow to wells, area
recharge, evapotranspiration, flow to drains, and flow through riverbeds can also
be simulated.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 53

Table 5. Continued

GWFM Description

A new flow and transport model, MODFLOW SURFACT, is based on the


USGS MODFLOW code, the most widely-used ground-water flow code in
the world. MODFLOW, however, has certain limitations in simulating
MODFLOW
complex field problems. Additional computational modules have been
SURFACT
incorporated to enhance the simulation capabilities and robustness.
MODFLOW SURFACT is a seamless integration of flow and transport
modules.

MODFLOWT is an enhanced version of the USGS MODFLOW model


which includes packages to simulate advective-dispersive contaminant
transport. Fully three-dimensional, MODFLOWT simulates transport of one
or more miscible species subject to adsorption and decay through advection
and dispersion. MODFLOWT performs groundwater simulations utilizing
transient transport with steady-state flow, transient flow, or successive
periods of steady-state flow. Groundwater flow data sets created for the
MODFLOWT original MODFLOW model function without alteration in MODFLOWT;
thus, extension of modeling projects to simulate contaminant transport is
very easy using MODFLOWT. It is thoroughly tested and has been bench-
marked against other transport codes including MT3D, SWIFT and
FTWORK. A comprehensive and pragmatic approach to contaminant
transport has been incorporated into MODFLOWT which allows for three
distinct directional dispersivity values, multiple chemicals and a rigorous
treatment of the hydrodynamic dispersion tensor.

MODFLOWwin32 has all the features of other MODFLOW versions


including the newest packages added over the years since MODFLOW's
original release by the USGS. These new packages include the Stream
Routing Package, Aquifer Compaction Package, Horizontal Flow Barrier
MODFLOWwin32 Package, BCF2 and BCF3 Packages, and the new PCG2 solver. In addition,
MODFLOWwin32 will create files for use with MODPATH (particle-
tracking model for MODFLOW) and MT3D (solute transport model).
MODFLOWwin32, as its name implies, is a 32-bit program designed to
address all the memory available to Windows.

MODPATH, "A Particle Tracking Post-Processing Package for MODFLOW,


MODPATH the USGS 3-D Finite-Difference Ground-Water Flow Model (MODFLOW)",
is a widely-used particle-tracking program.

ModTech is a new 3D groundwater flow and mass transport software.


ModTech enables you to model groundwater flow and contaminant transport
within a real-world geographic (GIS) environment. ModTech is more than
ModTech just another software package. Rather, as its name implies, it is a MODeling
TECHnology. It reaches deeply into the modeling concepts and comes up
with a variety of unique schematization approaches, which enable you to
solve some traditionally "unsolvable" problems.
54 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 5. Continued

GWFM Description
MOFAT for Windows includes a graphical preprocessor, mesh editor and
postprocessor with on-line help. Simulate multiphase (water, oil and gas)
flow and transport of up to five non-inert chemical species in MOFAT.
Model flow of light or dense organic liquids in three fluid phase systems.
Simulate dynamic or passive gas as a full three-phase flow problem. Model
water flow only, oil-water flow, or water-oil-gas flow in variably-saturated
porous media. MOFAT achieves a high degree of computational efficiency
by solving flow equations at each node (on the finite-element mesh) only for
phases that are undergoing changes in pressures and saturations above
MOFAT
specified tolerances using a new adaptive solution domain method.
Therefore, if NAPL is absent or exists at a residual saturation, MOFAT will
locally eliminate those flow equations. MOFAT analyzes convective-
dispersive transport in water, NAPL, and gas phases by assuming local
equilibrium or non-equilibrium partitioning among the fluid and solid phases.
MOFAT considers inter-phase mass transfer and compositional dependence
of phase densities. A concise but accurate description of soil capillary
pressure relations is used which assures natural continuity between single-
phase, two-phase and three-phase conditions.

MT3D is a comprehensive three-dimensional numerical model for simulating


solute transport in complex hydrogeologic settings. MT3D has a modular
design that permits simulation of transport processes independently or
jointly. MT3D is capable of modeling advection in complex steady-state and
transient flow fields, anisotropic dispersion, first-order decay and production
reactions, and linear and nonlinear sorption. It can also handle bioplume-type
MT3D
reactions, monad reactions, and daughter products. This enables MT3D to do
multi-species reactions and simulate or assess natural attenuation within a
contaminant plume. MT3D is linked with the USGS groundwater flow
simulator, MODFLOW, and is designed specifically to handle advectively-
dominated transport problems without the need to construct refined models
specifically for solute transport.

PEST is a nonlinear parameter estimation and optimization package. It can be


used to estimate parameters for just about any existing model whether or not
you have the model's source code. PEST is able to "take control" of a model,
PEST running it as many times as it needs while adjusting its parameters until the
discrepancies between selected model outputs and a complementary set of
field or laboratory measurements is reduced to a minimum in the weighted
least-squares sense.

PESTAN, Pesticide Transport, is a U.S. EPA program for evaluating the


transport of organic solutes through the vadose zone to groundwater.
PESTAN uses an analytical solution to calculate organic movement based on
PESTAN a linear isotherm, first-order degradation and hydrodynamic dispersion. Input
data includes water solubility, infiltration rate, bulk density, sorption
constant, degradation rates, saturated water content, characteristic curve
coefficient, saturated hydraulic conductivity and dispersion coefficient.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 55

Table 5. Continued

GWFM Description

Processing MODFLOW for Windows (PMWIN) is a complete simulation


system. It comes complete with a professional graphical preprocessor and
postprocessor, the 3-D finite-difference ground-water models MODFLOW-
Processing
88, MODFLOW-96, and MODFLOW 2000; the solute transport models
Modflow
MT3D, MT3DMS, RT3D and MOC3D; the particle tracking model
(PMWIN)
PMPATH 99; and the inverse models UCODE and PEST-ASP for automatic
calibration. A 3D visualization and animation package, 3D Groundwater
Explorer, is also included.

POLLUTE can be used for fast, accurate, and comprehensive contaminant


migration analysis. It implements a "1½-dimensional" solution to the
advection-dispersion equation. Unlike finite-element and finite-difference
formulations, POLLUTE does the numerical stability problems of alternate
approaches. Landfill designs that can be considered range from simple
systems on a natural clayey aquitard to composite liners with multiple
POLLUTE
barriers and multiple aquifers. In addition to advective-dispersive transport,
POLLUTE can consider adsorption, radioactive and biological decay, phase
changes, and transport through fractures. The Graphical User Interface makes
the editing, execution, and printing of data easy and flexible. This interface
includes also options to quickly design landfills with primary composite
barriers or primary and secondary composite barriers.

PRINCE is a well-known software package of ten analytical groundwater


models originally developed as part of an EPA 208 study. There are seven
one-, two- and three-dimensional mass transport models and three two-
dimensional flow models in PRINCE. These groundwater models have been
PRINCE rewritten from the original mainframe FORTRAN codes in graphics-rich and
PC-friendly C. Two popular analytical models have been added to the
original collection. The result is a widely acclaimed, user-friendly, menu-
driven package with built-in high resolution graphics for X-Y, 2-D contour
and 3-D surface plots.

The SEAWAT program was developed to simulate three-dimensional,


variable density, transient ground-water flow in porous media. The source
code for SEAWAT was developed by combining MODFLOW and
MT3DMS into a single program that solves the coupled flow and solute-
transport equations. The SEAWAT code follows a modular structure, and
thus, new capabilities can be added with only minor modifications to the
SEAWAT main program. SEAWAT reads and writes standard MODFLOW and
MT3DMS data sets, although some extra input may be required for some
SEAWAT simulations. This means that many of the existing pre- and post-
processors can be used to create input data sets and analyze simulation
results. Users familiar with MODFLOW and MT3DMS should have little
difficulty applying SEAWAT to problems of variable-density ground-water
flow.
56 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 5. Continued

GWFM Description
SESOIL is a seasonal compartment model which simulates long-term
pollutant fate and migration in the unsaturated soil zone. SESOIL describes
the following components of a user-specified soil column which extends
from the ground surface to the ground-water table: 1) Hydrologic cycle of the
unsaturated soil zone. 2) Pollutant concentrations and masses in water, soil,
and air phases. 3) Pollutant migration to groundwater. 4) Pollutant
volatilization at the ground surface. 5) Pollutant transport in washload due to
surface runoff and erosion at the ground surface. SESOIL estimates all of the
SESOIL above components on a monthly basis for up to 999 years of simulation time.
It can be used to estimate the average concentrations in groundwater. The
soil column may be composed of up to four layers, each layer having
different soil properties which affect the pollutant fate. In addition, each soil
layer may be subdivided into a maximum of 10 sublayers in order to provide
enhanced resolution of pollutant fate and migration in the soil column. The
following pollutant fate processes are accounted for: Volatilization,
Adsorption, Cation Exchange, Biodegradation, Hydrolysis and
Complexation.
SOLUTRANS is a 32-bit Windows program for modeling three-dimensional
solute transport based on the solutions presented by Leij et al. for both
equilibrium and non-equilbrium transport. The interface and input
requirements are so simple that it only takes a few minutes to develop models
SOLUTRANS and build insight about complex solute transport problems. With
SOLUTRANS you can, in a matter of minutes, model solute transport from a
variety of source configurations and build important insights about key
processes. SOLUTRANS offers a quick and simple alternative to complex,
time-consuming 3-D numerical flow and transport models.
SUTRA is a 2D/3D groundwater saturated-unsaturated transport model, a
complete saltwater intrusion and energy transport model. SUTRA simulates
fluid movement and transport of either energy or dissolved substances in a
subsurface environment. SUTRA employs a two-dimensional hybrid finite-
element and integrated finite-difference method to approximate the
SUTRA governing equations that describe the two interdependent processes that are
simulated: (1) fluid densitydependent saturated or unsaturated groundwater
flow and either (2a) transport of a solute in the groundwater, in which the
solute may be subject to equilibrium adsorption on the porous matrix and
both first-order and zero-order production or decay, or (2b) transport of
thermal energy in the groundwater and solid matrix of the aquifer.
SVFlux 2D represents the next level in seepage analysis software. Designed
to be simple and effective, it offers features designed to allow the user to
focus on seepage solutions, not convergence problems or difficult mesh
creation. Great care has been taken to model geometry CAD-style input after
SVFlux 2D the popular AutoCAD(TM) software. Freeform boundary equations and an
optional soil database of over 6,000 soils to choose from further simplify
model design. The finite element solution makes use of fully automatic mesh
generation and mesh refinement to solve the problem quickly as well as
indicating zones of critical gradient.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 57

Table 5. Continued

GWFM Description
SWIFT is a fully-transient, three-dimensional model to simulate groundwater
flow, heat (energy), brine and radionuclide transport in porous and fractured
geologic media. The primary equations for fluid (flow), heat and brine are
coupled by fluid density, viscosity and porosity. In addition to transient
analysis, SWIFT offers a steady-state option for coupled flow and brine. The
equations are solved using central or backward spatial and time weighting
SWIFT
approximations by the finite-difference method. In addition to Cartesian,
cylindrical grids may be used. Contaminant transport includes advection,
dispersion, sorption and decay, including chains of constituents. Both dual-
porosity and discrete-fracture representations along with rock matrix
interactions may be simulated. The nonlinearities resulting from water table
and variable density are solved iteratively.
SWIMv1 (Soil Water Infiltration and Movement model version 1) is a
software package for simulating water infiltration and movement in soils.
SWIMv1 consists of a menu-driven suite of three programs that allow the
user to simulate soil water balances using numerical solutions of the basic
soil water flow equations. As in the real world, SWIMv1 allows addition of
water to the system as precipitation and removal by runoff, drainage,
evaporation from the soil surface and transpiration by vegetation. SWIMv1
helps researchers and consultants understand the soil water balance so they
can assess possible effects of such practices as tree clearing, strip mining and
irrigation management. SWIMv1 is valuable for scientists and consultants
involved in land planning and land management. SWIMv2 (Soil Water
SWIMv1/SWIMv2
Infiltration and Movement model version 2) is a mechanistically-based
model designed to address soil water and solute balance issues associated
with both production and the environmental consequences of production.
SWIMv2 employs fast, numerically-efficient techniques for solving
Richards' equation for water flow and the convection-dispersion equation for
solute transport and is suitable for personal computer applications. The
model deals with a one-dimensional vertical soil profile which may be
vertically inhomogeneous but is assumed to be horizontally uniform. It can
be used to simulate runoff, infiltration, redistribution, solute transport and
redistribution of solutes, plant uptake and transpiration, evaporation, deep
drainage and leaching.
TWODAN is a popular and versatile analytic ground-water flow model for
Windows. TWODAN has a suite of advanced analytic modeling features that
allow you to model everything from a single well in a uniform flow field to
complex remediation schemes with numerous wells, barriers, surface waters,
and heterogeneities. TWODAN has many capabilities: heterogeneities,
impermeable barriers, resistant barriers, and transient solutions, to name a
TWODAN few. The analytic method of TWODAN demands minimal input and the new
seamless Windows interface can be quickly mastered. TWODAN can be
used for modeling remedial design alternatives, wellhead capture zones, and
regional aquifer flow. TWODAN combines advanced analytic elements with
a user interface. Comparing this software with other similar GWFM, it is a
good tool for most remediation design, capture zone analysis, and regional
modeling problems.
58 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Table 5. Continued

GWFM Description
Visual Groundwater is a 3-D visualization software package which can be
used to deliver high-quality, three-dimensional presentations of subsurface
characterization data and groundwater modeling results. Visual Groundwater
combines state-of-the-art graphical tools for 3-D visualization and animation
Visual
with a data management system specifically designed for borehole
Groundwater
investigation data. Visual Groundwater also comes with a data conversion
utility to create 3-D data files using random X, Y, Z data and gridded data
sets. Three-dimensional images of complex site characterization data and
modeling results can be easily created.
Visual HELP for Windows 95/98/2000/NT is an advanced hydrological
modeling environment available for designing landfills, predicting leachate
mounding and evaluating potential leachate contamination. Visual HELP
combines the latest version of the HELP model with an easy-to-user interface
and powerful graphical features for designing the model and evaluating the
modeling results. Visual HELP's user-friendly interface and flexible data
VISUAL HELP handling procedures provides convenient access to both the basic and
advanced features of the HELP model. This completely-integrated modeling
HELP environment allows the user to graphically create several profiles
representing different parts of a landfill; automatically generate statistically-
reliable weather data (or create your own); run complex model simulations;
visualize full-color, high-resolution results; and prepare graphical and
document materials for your report.

Visual MODFLOW provides professional 3D groundwater flow and


contaminant transport modeling using MODFLOW-2000, MODPATH,
MT3DMS and RT3D. Visual MODFLOW Pro seamlessly combines the
standard Visual MODFLOW package with WinPEST and the Visual
MODFLOW 3D-Explorer to give the most complete and powerful graphical
Visual modeling environment available. This fully-integrated groundwater modeling
MODFLOW environment allows to: 1) Graphically design the model grid, properties and
boundary conditions, 2) Visualize the model input parameters in two or three
dimensions, 3) Run the groundwater flow, path-line and contaminant
transport simulations, 4) Automatically calibrate the model using WinPEST
or manual methods, and 5) Display and interpret the modeling results in
three-dimensional space using the Visual MODFLOW 3D-Explorer
Visual PEST combines the powerful parameter estimation capabilities of
PEST2000 with the graphical processing and display features of WinPEST.
PEST2000 is the latest version of PEST, the pioneer in model-independent
parameter estimation. Since it was first released over six years ago, PEST has
gained extensive use all over the world in many different fields. During this
VISUAL PEST time it has undergone continued development with the addition of many new
features that have improved its performance and utility to a level that makes
it uniquely applicable in just about any modeling environment. PEST is now
used extensively for automated model calibration and data interpretation in
groundwater and surface-water hydrology, geophysics, geotechnical,
mechanical and mining engineering as well as many other fields.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 59

Table 5. Continued

GWFM Description
VS2DT is a USGS program for flow and solute transport in variably-
saturated, single-phase flow in porous media. A finite-difference
approximation is used in VS2DT to solve the advection-dispersion equation.
Simulated regions include one-dimensional columns, two-dimensional
vertical cross sections, and axiallysymmetric, three-dimensional cylinders.
The VS2DT program options include backward or centered approximations
for both space and time derivatives, first order decay, equilibrium adsorption
VS2DT
(Freundlich or Langmuir) isotherms, and ion exchange. Nonlinear storage
terms are linearized by an implicit Newton-Raphson method. Relative
hydraulic conductivity in VS2DT is evaluated at cell boundaries using full
upstream weighting, arithmetic mean or geometric mean. Saturated hydraulic
conductivities in VS2DT are evaluated at cell boundaries using distance-
weighted harmonic means. A graphical user interface for VS2DT is now
available in the WHI UnSat Suite.
WinFlow is a powerful yet easy-to-use groundwater flow model. WinFlow is
similar to Geraghty & Miller's popular QuickFlow model which was
developed by one of the authors of QuickFlow. WinFlow is a true Windows
program incorporating a multiple document interface (MDI). WinFlow is an
interactive analytical model that simulates two-dimensional steady-state and
WinFlow transient groundwater flow. The steady-state module in WinFlow simulates
groundwater flow in a horizontal plane using analytical functions developed
by Strack. The transient module uses equations developed by Theis and by
Hantush and Jacob for confined and leaky aquifers, respectively. Each
module uses the principle of superposition to evaluate the effects from
multiple analytical functions (wells, etc.) in a uniform regional flow field.
WinTran couples the steady-state groundwater flow model from WinFlow
with a contaminant transport model. The transport model has the feel of an
analytic model but is actually an embedded finite-element simulator. The
finite-element transport model is constructed automatically by WinTran but
displays numerical criteria (Peclet and Courant numbers) to allow the user to
avoid numerical or mass balance problems. Contaminant mass may be
injected or extracted using any of the analytic elements including wells,
WinTran ponds, and linesinks. In addition, constant concentration elements have been
added. WinTran displays both head and concentration contours, and
concentration may be plotted versus time at selected monitoring locations.
The transport model includes the effects of dispersion, linear sorption
(retardation), and first-order decay. WinTran aids in risk assessment
calculations by displaying the concentration over time at receptor or
observation locations. WinTran will display the breakthrough curves after the
simulation is finished.

Considering the large variability and the quick development of groundwater models, a
new, more sophisticated model can often replace a previously applied model. Additionally,
the reconsideration of the conceptual model and the regeneration of the mesh may need a new
allocation of the parameters. Therefore, it is important that model data (information) are
stored independently from a given model, with a preference for GIS-based databases.
60 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Considerable development in the field of user-friendly GIS and data base servers make the
set-up and the modification of models easier and more time effective. One such model is
GroundWaterVistas which incorporates mathematical modeling with GIS-based data
exchange interfaces.
Data used in groundwater modeling consist of four categories: (1) the aquifer-system
stress factors, (2) the aquifer-system and strata geometry, (3) the hydrogeological parameters
of the simulated process; and (4) the main measured variables. Stress factors for groundwater
flow include: effective recharge, pumping volumes, water-surface flow exchanges, etc. In
contaminant-transport modeling, the input and output contaminant mass flows are stress
factors. These stress factors are imposed on the model through the “boundary” conditions
(Franke et al. 1987; ASTM 5609-94) or “source/sink” terms. Appropriate aquifer-system
geometry can be determined using geological information (maps and cross sections),
topographic maps, and contour maps of the upper and lower limits for the aquifer strata.
Initial estimates of the distributed values and spatial distributions of the hydrogeological
parameters (hydraulic conductivity, storage coefficient, dispersivity, etc.) need to be made
using raw data and interpretations.
Of course, the interpretation is based on knowledge of the aquifer geology and
hydrogeology. Maps and cross sections representing the spatial variations of hydrogeological
parameter values are used. For a flow problem, the main measured variable is the hydraulic
head, and for a contaminant-transport problem, it is the contaminant concentration. These
consist of point values measured at different time periods in the entire aquifer. They are
required for model calibration and validation. Links can be organized between models and
GIS using three techniques: loose coupling, tight coupling, and embedded coupling. Loose
coupling is when the GIS and the model represent distinct software packages and the data
transfer is made through input/output model-predefined files. The GIS software is used to pre-
process and post-process the spatial data. An advantage of this solution is that the coupled
software packages are independent systems, facilitating potential future changes in an
independent manner. In tight coupling, an export of data to the model from GIS is performed,
but the GIS tools can interactively access input model subroutines. In this case, the data
exchange is fully automatic. When a model is created using the GIS programming language
or when a simple GIS is assimilated by a complex modeling system, embedded coupling is
used. Tight coupling as well as embedded coupling involves a significant investment in
programming and data management that is not always justified.

Application of GWFM in Keritis River Basin

Except the previously mentioned collection and manipulation of the available data for the
Keritis basin (Kouli et al. 2005; Kouli et al. 2007a;b) several electrical geophysical
measurements (VES) were acquired in the study area and,in conjunction with the above
mentioned hydrogeological (groundwater depth, aquifer thickness, etc.) information as
derived from the existing boreholes, gave an accurate estimation of the formation factor and
the hydraulic parameters of the aquifer (transmissivity, hydraulic conductivity and transverse
resistance) (Soupios et al. 2007).
After all the aforementioned data collection / processing / manipulation the application of
the GV software was initiated (Soupios and Stavroulakis 2006). First, the finite difference
grid (regular mesh) network (i.e., the number of rows, columns, and layers, the number of
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 61

stress periods, row, column, and layer spacings, and default values for all properties) was
designed. Then, a base map (ArcView shapefile format) with the model grid was imported.
Specifically, the shapefiles were imported with attributes to set boundary conditions (BC) and
the aquifer properties (hydraulic and transport). The BCs were assigned using head-dependent
boundary condition in rivers, time-varying head, where the head should be constant for a
particular time step, but the head may change between time steps and constant flux boundary

Figure 10. The initial model was created and the border of the studied river basin (Keritis Basin) was
also imported. The drainage system (as derived from GIS), the most representative hydrowells (H) of
the study area and the circular boundary conditions (shown with blue circles) in each of the boreholes
are also included (Soupios and Stavroulakis 2006).
62 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

for nine representative hydrowells in the study area. For each of the boreholes the circular
boundary was also designed. The next process step was the assignment of the aquifer
(hydraulic and transport) properties. Some of the required properties were estimated from
Soupios et al. (2007) and entered into the database. Finally, the model design was saved and it
was ready for running the simulation (Figures 10, 11).

Figure 11. The study area was classified in four hydraulic zones, based on previous studies (Soupios et
al. 2007) and the known hydrolithological and permeability characteristics of the area (Soupios and
Stavroulakis 2006).
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 63

Running model simulations involves three steps which follow the design of the groundwater
model. Thus, the specific options for the particular model in use (e.g., MODFLOW) were
edited, the input files for the model were created and finally the model was run.

Figure 12. The contours of the head results for the current layer and cross-section views are presented.
The double red line depicts the starting location for the particle tracking using MODPATH and
PATH3D (Soupios and Stavroulakis 2006).

The pre- and post-processing of particle tracking analysis (transport modeling), (Shafer
1987; Pollack 1988, 1989; Zheng 1991), was also applied using the MODPATH and for
PATH3D algorithms (Figure 12) (Konikow and Grove 1977; Anderson 1979; Wang and
Anderson 1982; Bear and Verruijt 1987; Reilly et al. 1987; Zheng 1990; Anderson and
Woessner 1992; ASTM 978-92; ASTM 5880-95; Zheng and Bennett 1995; Zheng et al.
64 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

2001). We should note that particle tracking neglects the effects of chemical reactions,
dispersion, and diffusion. The simulation results of the particle tracking are displayed by
plotting pathlines through the aquifer system (Figure 13).

Figure 13. The contours of the head results for the current layer and cross-section views are presented.
The results of a particle tracking simulation are displayed by plotting path lines through the aquifer
system. Travel times are also labelled on the path lines (Soupios and Stavroulakis 2006).

Conclusions
In nowadays, the vitally important groundwater resources undergo severe risk due to the rapid
increase of population, modern land use applications (agricultural and industrial) and growing
demands for water supply. Moreover, Southern Europe, and particularly the Mediterranean
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 65

region, is extremely prone to groundwater problems as it suffers from long dry periods.
Assessing and monitoring the quality of groundwater is therefore important to assure the
sustainable use of these resources.
More and more GIS-based applied groundwater research is required in conjunction with
field investigations to effectively exploit the expanding potential of GIS technologies, which
will perfect and standardize current applications as well as evolve new approaches and
applications in the future.
The guidance and directions given in this chapter will provide a framework for managing
groundwater resources through collection, management, and analysis of data; conceptual and
numerical modeling of aquifers; and evaluation of aquifer yield. Evaluation of the
consequences of management options can also be examined based on this framework.
The integrated use of GIS and groundwater models will produce a series of models and
databases which will be maintained and used in assessing management options. Moreover, it
will provide a framework for model updates and additional studies, and it will contribute to
optimization of the research processes in different stages and to amalgamate data from
various sources.
However, errors and uncertainties in a groundwater flow analysis and solute transport
analysis make any model prediction no better than an approximation. Moreover, precisely
depicting the water quality conditions is difficult due to the spatial and temporal variability of
multiple contaminants and to the wide range of indicators (chemical, physical and biological)
that might be measured.
For this reason, all model predictions should be expressed as a range of possible
outcomes that reflect the assumptions involved and uncertainties (modeling errors) in model
input data and parameter values. The best method of eliminating or reducing modeling errors
is to apply good hydrogeological judgment and to question the model simulation results. If
the results do not make physical sense, find out why.

Acknowledgments

The project is co-funded by the European Social Fund and National Resources in the
framework of the project INTERREG III B ARCHIMED, sub-project A1.020 entitled
“MILDMAP - Methodology Integration of EO techniques as operative tool for Land
Degradation Management and planning in Mediterranean Areas”. NLS and PS acknowledge
co-funding by the European Social Fund and National Resources, EPEAEK II –
ARCHIMEDES. The authors are grateful to Dr. A. Sarris and Dr. G. Vargemezis for fruitful
discussions, helpful comments and valuable suggestions.

References
Adams B. and Foster S. (1992) Land-surface zoning for groundwater protection. J Inst Water
Environ Manage, 6, pp.312–320.
Al-Adamat R.A.N., Foster I.D.L. and Baban S.M.J. (2003) Groundwater vulnérabilité and
risk mapping for the Basaltic aquifer of the Azraq basin of Jordan using GIS, Remote
sensing and DRASTIC, Applied Geography, 23, pp.303–324.
66 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Albinet M. and Margat J. (1970) Cartographie de la Vulnérabilité la Pollution des Nappes


d'eau Souterraine Orleans, Fr. Bull. BRGM 2_me Srie, 4, pp.13–22.
Aller L., Bennett T., Lehr J.H., Petty R.J. and Hackett G. (1987) DRASTIC: a standardized
system for evaluating ground water pollution potential using hydrogeologic settings,
United States Environ Prot Agency, Ada, Oklahoma 74820, EPA 600/2-85/018, 455p.
American Society for Testing and Materials (ASTM), Standard Guide for Application of a
Ground-Water Flow Model to a Site-Specific Problem. ASTM Standard D 5447-93, 6 p.
American Society for Testing and Materials (ASTM), Standard Guide for Comparing
Ground-Water Flow Model Simulations to Site Specific Information. ASTM Standard D
5490-93, 7 p.
American Society for Testing and Materials (ASTM), Standard Guide for Defining Boundary
Conditions in Ground-Water Flow Modeling. ASTM Standard D 5609-94, 4 p.
American Society for Testing and Materials (ASTM), Standard Guide for Defining Initial
Conditions in Ground-Water Flow Modeling. ASTM Standard D 5610-94, 2 p.
American Society for Testing and Materials (ASTM), Standard Guide for Conducting a
Sensitivity Analysis for a Ground-Water Flow Model Application. ASTM Standard D
5611-94, 5 p.
American Society for Testing and Materials (ASTM), Standard Guide for Subsurface Flow
and Transport Modeling. ASTM Standard D 5880-95, 6 p.
American Society for Testing and Materials (ASTM), Standard Guide for Developing
Conceptual Site Models for Contaminated Sites. ASTM Standard E 1689-95, 8 p.
American Society for Testing and Materials (ASTM), Standard Guide for Subsurface Flow
and Transport Modeling. ASTM Standard D 5880-95, 6 p.
American Society for Testing and Materials (ASTM), Standard Guide for Calibrating a
Ground-Water Flow Model Application. ASTM Standard D 5918-96, 6 p.
American Society for Testing and Materials (ASTM), Standard Practice for Evaluating
Mathematical Models for the Environmental Fate of Chemicals. ASTM Standard E 978-
92, 8 p.
Anderson M.P. and Woessner W.W. (1992) Applied Groundwater Modeling. Academic
Press, Inc., San Diego, CA., 381 p.
Anderson M.P. (1979) Using models to simulate the movement of contaminants through
groundwater flow systems. CRC Critical Review in Environmental Control, No. 9, pp.
97-156.
Anderson M.P. (1984) Movement of Contaminants in Groundwater: Groundwater Transport-
Advection and Dispersion: in Groundwater Contamination, Studies in Geophysics.
National Academy Press, Washington D.C., pp. 429-437.
Andreo B., Goldscheider N., Vadillo I., Vias J., Neukum C., Sinreich M., Jimenez P.,
Brechenmacher J., Carrasco F., Hötzl H. and Perles M.J. (2006) Karst groundwater
protection: first application of a Pan-European approach to vulnerability, hazard and risk
mapping in the Sierra de Libar (southern Spain). Sci Total Environ 357, pp.54–73.
Argialas D., Stearns R. and Shahrokhi F. (1988) Mapping and significance of Landsat and
gravity lineaments in west Tennessee: Jour. Aerospace Engineering, Am. Soc. Civil
Engineers, 1(2), pp.74–87.
Aronoff S. (1989) Geographic Information Systems: A Management Perspective, WDL
Publications, Ottawa.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 67

APHA, American Public Health Association, (2005) Standard Methods for the Examination
of Water and Wastewater, 21st ed (centennial edition), APHA.
Babiker I.S., Mohamed M.A.A., Terao H., Kato K. and Ohta K. (2004) Assessment of
groundwater contamination by nitrate leaching from intensive vegetable cultivation using
geographical information system. Environ Int., 29(8) pp.1009– 17.
Babiker I.S., Mohamed M.A.A., Hiyama T. and Kato K. (2005) A GIS-based DRASTIC
model for assessing aquifer vulnerability in Kakamigahara Heights, Gifu Prefecture,
central Japan, Science of the Total Environment, 345, pp.127–140.
Babiker I.S., Mohamed M.A.A. and Hiyama T. (2007) Assessing groundwater quality using
GIS, Water Resour Manage, 21, pp.699–715
Barbash J.E. and Resek E.A. (1996) Pesticides in ground water: distribution, trends, and
governing factors. Chelsea, MI7 Ann Arbor Press, 1996.
Barrocu G., Muzzu M. and Uras G. (2007) Hydrogeology and vulnerability map (Epik
method) of the ‘‘Supramonte’’ karstic system, north-central Sardinia, Environ Geol, 51,
pp. 701–706.
Bear J. and Verruijt A. (1987) Modeling Groundwater Flow and Pollution. D. Reidel
Publishing Company, 414 p.
Bennett G.D. (1976) Introduction to Ground-Water Hydraulics - A Programmed Text for
Self-Instruction, Techniques of Water-Resources Investigations of the United States
Geological Survey, Book 3, Chapter B2, 172 p.
Biedny D. and Monroe B. (1991) Official photoshop handbook: Bantam, New York, p. 423.
Bobba A.G., Bukata R.P. and Jerome J.H. (1992) Digitally processed satellite data as a tool in
detecting potential groundwater flow systems, Journal of Hydrology 131(1-4), pp.25–62.
Bonham-Carter G.F. (1996) Geographic information systems for geoscientists: modeling with
GIS computer methods in the geosciences, 13, Elsevier Science Ltd, Pergamon, pp.1–50.
Brechenmacher J. (2002) Application of the PI method for groundwater vulnerability
mapping in the karst aquifer of Sierra de Lìbar (Andalusia/Spain). Processing and
analysis with GIS. Diplom. Thesis in Geoecology, University of Karlsruhe, unpublished.
Bredehoeft J.D. and Konikow L.F. (1993) Ground-Water Models: Validate or Invalidate.
Ground Water, Vol. 31, No. 2, p. 178-179.
Bredehoeft J.D. and Pinder G.F. (1973) Mass Transport in Flowing Groundwater, Water
Resources Research, 9, pp.194-210.
Brito M.G., Costa C.N., Almeida J.A., Vendas D. and Verdial P.H. (2006) Characterization of
maximum infiltration areas using GIS tools Engineering Geology 85, pp.14–18.
Brouyère S., Jeannin P.Y., Dassargues A., Goldscheider N., Popescu C., Sauter M., Vadillo I.
and Zwahlen F. (2001) Evaluation and validation of vulnerability concepts using a
physically based approach, 7th Conference on Limestone Hydrology and Fissured Media,
Besan¸ on 20–22 September 2001. Sci Tech Environm Mèm, 13, pp.67–72.
Brouyère S. (2004) A quantitative point of view of the concept of vulnerability. In: Zwahlen
F (ed) Vulnerability and risk mapping for the protection of carbonate (karst) aquifers,
final report COST Action 620. European Commission Directorate-General for Research,
EUR 2091:10–15.
Burrough P.A. (1986) Principles of Geographical Information Systems for Land Resources
Assessment, Oxford University Press, Oxford, pp.193.
Burrough P. and McDonell R. (1998) Principles of geographical information systems. Oxford
University Press.
68 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Carter A.D., Palmer R.C. and Monkhouse R.A. (1987) Mapping the vulnerability of
groundwater to pollution from agricultural practice, particularly with respect to nitrate.
In: Duijvenbooden, W., Waegeninghm HG. (eds) Vulnerability of soil and groundwater
to pollutants. TNO Committee on Hydrological Research, The Hague, Proc Info, 38,
pp.333–342
Chartzoulakis K.S., Paranychianakis N.V. and Angelakis A.N. (2001) Water resources
management in the island of Crete, Greece, with emphasis on the agricultural use,
WaterPolicy, 3, pp.193-205.
Chatterjee C., Rakesh Kumar R., Chakravorty B., Lohani A.K. and Kumar S. (2005)
Integrating Remote Sensing and GIS Techniques with Groundwater Flow Modeling for
Assessment of Waterlogged, Water Resources Management, 19(5), DOI 10.1007/s11269-
005-2071-4.
Cherry J.A., Gillham R.W. and Barker J.F. (1984) Contaminants in Groundwater: Chemical
Process, in Groundwater Contamination, Studies in Geophysics. National Academy
Press, Washington D.C., pp. 46-63.
Chowdary V.M., Rao N.H. and Sarma P.B.S. (2003) GIS-based decision support system for
groundwater assessment in large irrigation project areas. Agric. Water Manage. 62,
pp.229–252.
Civita M. (1994) Le carte della vulnerabilità degli acquiferi all’inquinamiento: teoria e pratica
[Contamination vulnerability mapping of the aquifer: theory and practice]. Quaderni di
Tecniche di Protezione Ambientale, Pitagora Editrice.
Civita M. and De Maio M. (2000) Valutazione e cartografia automatica della vulnerabilità
degli acquiferi all’inquinamento con il systema parametrico SINTACS R5 [Evaluation
and automatic cartography of aquifer vulnerability using the parametric system
SINTACS R5]. Pitagora Editrice, Bologna.
Clarke K. (2001) Getting Started with Geographic Information Systems, 3rd edition, Prentice
Hall, New Jersey -USA.
Commission on Geosciences Environment, and Resources. National Academy Press,
Washington, DC, pp 204.
Cooley R.L. (1992) A MODular Finite-Element model (MODFE) for areal and axisymmetric
ground-water-flow problems, part 2: derivation of finite-element equations and
comparisons with analytical solutions: U.S. Geological Survey Techniques of Water-
Resources Investigations, Book 6, Chapter A4, 108 p.
Corniello, A. and Ducci, D. (2000) Pollution vulnerability assessment in karstic aquifers. A
case study of the Matese Mountains. (Southern Italy). In: O. Sililo et al. eds.:
Groundwater: Past Achievements and Future Challenges, pp. 725-730, Balkema,
Rotterdam.
Daly D., Dassargues A., Drew D., Dunne S., Goldscheider N., Neale S., Popescu C. and
Zwhalen F. (2002) Main concepts of the “European Approach” for (karst) groundwater
vulnerability assessment and mapping, Hydrogeol J, 10(2), pp.340–345.
Deckers F. and Te Stroet C.B.M. (1996) Use of GIS and database with distributed modeling.
In: Abbott MB, Refsgaard JC (eds), Distributed Hydrological Modeling, Kluwer
Academic Publishers, Dordrecht, pp. 215–232.
DeMers MN. (2000) Fundamentals of Geographic Information Systems, 2nd edition,
JohnWiley & Sons, Inc., New York
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 69

DeVantier B.A. and Feldman A.D. (1993) Review of GIS applications in hydrologic
modeling. Journal of Water Resources Planning and Management, ASCE, 119(2),
pp.246–261
DoELG/EPA/GSI (1999) Groundwater protection schemes. Department of Environment and
Local Government, Environmental Protection Agency and Geological Survey of Ireland,
Dublin, 24 pp.
Doerfliger N. and Zwahlen F. (1995) EPIK: a new method for outlining of protection areas: a
water vulnerability assessment in karst environment. In: Proceedings of 5th International
Symposium on karst waters and environmental impacts, Antalya, Balkema, Rotterdam
pp.117–123.
Doerfliger N. and Zwahlen F. (1998) Groundwater Vulnerability Mapping in Karstic Regions
(EPIK) – Application to Groundwater Protection Zones. Swiss Agency for the
Environment, Forests and Landscape (SAEFL), Bern
Doerfliger N., Jeannin P.Y. and Zwahlen F. (1999) Water vulnerability assessment in karst
environments: a new method of defining protection areas using a multi-attribute approach
and GIS tools (EPIK method). Environ Geol, 39(2), pp.165–176.
Drew D. and Hotzl H. (eds) (1999) Karst Hydrogeology and Human Activities. Impacts,
Consequences and Implications. International Contributions to Hydrogeology (IAH), 20.
Balkema, Rotterdam
Durnford D.S., Thompson K.R., Ellerbrook D.A., Loftis J.C. and Davies G.S. (1990)
Screening methods for ground water pollution potential from pesticide use in Colorado
agriculture. Completion Report Colorado Water Resources Research Institute, Fort
Collins, pp. 157–165.
EC (European Community), European Community Council Directive 98/83/EC (1998) on the
quality of water intended for human consumption. Official Journal of the European
Communities
Engman E.T. and Gurney R.J. (1991) Remote Sensing in Hydrology. Chapman and Hall,
London, 225 pp
Evans B.M. and Myers W.L. (1990) A GIS-based approach to evaluating regional
groundwater pollution potential with DRASTIC, Journal of Soil and Water Conservation,
45, pp.242–245.
Fabbri A.G. and Napolitano P. (1995) The use of database management and geographical
information systems for aquifer vulnerability analysis. Contribution to the International
Scientific Conference on the occasion of the 50th Anniversary of the founding of the
Vysoka Skola Banska, Ostrava, Czech Republic.
Farnsworth R.K., Barret E.C. and Dhanju M.S. (1984) Application of Remote Sensing to
Hydrology including Ground Water. IHP-II Project A. 1.5, UNESCO, Paris
Faust N., Anderson W.H. and Star J.L. (1991) Geographic information systems and remote
sensing future computing environment. Photogrammetric Engineering and Remote
Sensing, 57(6), pp.655–668
Ford D. and Williams D.W. (1989) Karst geomorphology and hydrology. Unwin Hyman,
Boston, 601 pp.
Foster I.D.L. and Ilbery B.W. (1992) Water protection zones: A valid management strategy?
In A. W. Gilg (Ed.), Restructuring the countryside, London:Avebury Studies in Green
Research, pp.203–222.
70 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Foster I.D.L. and Thorn R.H. (1993) Contrasting problems of implementing European
Community Water Protection Zone Policies in Ireland and Great Britain. Jnl. Inst. Water
& Environ. Managers, 7, pp.62–67.
Foster S. and Hirata R. (1988) Groundwater pollution risk assessment: a methodology using
available data, WHO-PAHO-CEPIS, Lima
Foster S.S.D. (1987) Fundamental concepts in aquifer vulnerability, pollution risk and
protection strategy. In: Duijvenbooden W, Waegeningh HG (eds) Vulnerability of soil
and groundwater to pollutants. TNO Committee on Hydrological Research, The Hague,
Proc Info, 38, pp.69–86.
Francés A., Paralta E., Fernandes J. and Ribeiro L. (2001) Development and application in the
Alentejo region of a method to assess the vulnerability of groundwater to diffuse
agricultural pollution: the susceptibility index, 3rd International Conference on Future
Groundwater Resources at Risk, Lisbon (Portugal), 25–27 June (2001).
Franke O.L., Bennett G.D., Reilly T.E., Laney R.L., Buxton H.T. and Sun R.J. (1991)
Concepts and Modeling in Ground-Water Hydrology-A Self-Paced Training Course. U.S.
Geological Survey Open-File Report 90-707.
Franke O.L., Reilly T.E. and Bennett G.D. (1987) Definition of Boundary and Initial
Conditions in the Analysis of Saturated Ground-Water Flow Systems - An Introduction.
U.S. Geological Survey Techniques of Water-Resources Investigations, Book 3, Chapter
B5, 15 p.
Freeze R.A. and Witherspoon P.A. (1966) Theoretical Analysis of Regional Groundwater
Flow: 1. Analytical and Numerical Solutions to the Mathematical Model, Water
Resources Research, 2(4), pp.641-656.
Freeze R.A. and Witherspoon P.A. (1967) Theoretical Analysis of Regional Groundwater
Flow: 2. Effect of Water-Table Configuration and Subsurface Permeability Variation,
Water Resources Research, 3(2), pp.623-634.
Fritch T.G., McKnight C.L., Yelderman Jr J.C. and Arnold J.G. (2000) An aquifer
vulnerability assessment of the paluxy aquifer, central Texas, USA, using GIS and a
modified DRASTIC approach, Environmental Management, 25, pp.337–345.
Fryberg D.L. (1988) An Exercise in Ground-Water Model Calibration and Prediction, Ground
Water, 26(3), pp.350-360.
Ghosh N.C. and Sharma K.D. (2006) Groundwater Modelling and Management, New Delhi,
Capital Pub., pp.596, tables, figs., ISBN 81-85589-44-5.
Gleeson T.A. (1967) On Theoretical Limits of Predictability. Journal of Meteorology, 6(2),
p.213-215.
Gogu R.C. and Dassargues A. (2000) Current Trends and Future Challenges in Groundwater
Vulnerability Assessment using Overlay and Index Methods, Environ. Geol., 39(6),
pp.549 – 559.
Gogu R., Carabin G., Hallet V., Peters V. and Dassargues A. (2001) GIS-based
hydrogeological databases and groundwater modelling, Hydrogeology Journal, 9(6), pp.
555-569, doi 10.1007/s10040-001-0167-3.
Goldscheider N., Klute M., Sturm S. and Hotzl H. (2000) The PI method: a GIS-based
approach to mapping groundwater vulnerability with special consideration of karst
aquifers, Z Angew Geol, 463, pp.157–166.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 71

Goldscheider N., Hotzl H., Fries W. and Jordan P. (2001) Validation of a vulnerability map
(EPIK) with tracer tests. 7th Conference on Limestone Hydrology and Fissured Media,
Besancon 20–22 September 2001, Sci. Tech. Environm. Mem., 13, pp.167–170.
Goldscheider N. (2004) The concept of groundwater vulnerability. In: Zwahlen F (ed)
Vulnerability and risk mapping for the protection of carbonate (karst) aquifers, final
report COST Action 620. European Commission, Directorate-General for Research, EUR
20912:5–9
Goldscheider N. and Popescu I.C. (2004) The European approach. In: Zwahlen F (ed)
Vulnerability and risk mapping for the protection of carbonate (karst) aquifers, final
report COST Action 620. European Commission Directorate-General for Research, EUR
20912:17–21.
Goodchild M.F. (1992) Geographical information science. International Journal of
Geographical Information Systems, 6(1), pp.31–45.
Goodchild M.F. (1993) The state of GIS for environmental problem-solving. In: Goodchild
MF, Parks BO, Steyaert LT (eds), Environmental Modeling with GIS, Oxford University
Press, New York, pp. 8–15.
Goodchild M.F., Parks B.O. and Steyaert L.T. (1993) Environmental Modeling with GIS,
Oxford University Press US, pp. 520, ISBN 0195080076.
Gossel W., Ebraheem A.M. and Wycisk P. (2004) A very large scale GIS-based groundwater
flow model for the Nubian sandstone aquifer in Eastern Sahara (Egypt, northern Sudan
and eastern Libya), Hydrogeology Journal, 12(6), pp.698–713.
Gupta R. (1991) Remote sensing geology: Springer–Verlag, Berlin, 233 p.
Halliday S.L. and Wolfe M.L. (1991) Assessing ground water pollution potential from
nitrogen fertilizer using a geographic information system, Water Resour. Bull., 27,
pp.237–245.
Hanson R.T. and Leake S.A. (1999) Documentation of HYDMOD, a program for extracting
and processing time-series data from the U.S. Geological Survey's modular three-
dimensional finite-difference ground-water flow model: U.S. Geological Survey Open-
File Report 98-564, 57 p.
Haralick R.M. (1984) Digital step edges from zero crossing of second directional filters:
IEEE Trans. Pattern Analysis and Machine Intelligence, PAMI-6, p. 58–68.
Harbaugh A.W., Banta E.R., Hill M.C. and McDonald M.G. (2000) MODFLOW-2000, the
U.S. Geological Survey modular ground-water model - User guide to modularization
concepts and the Ground-Water Flow Process: U.S. Geological Survey Open-File Report
00-92, 121 p.
Heilman J.L. and Moore D.G. (1981) Groundwater applications of the Heat Capacity
Mapping Mission. Satellite Hydrology, AWRA, Minneapolis, MN pp.446–449.
Hill M.C. (1998) Methods and Guidelines for Effective Model Calibration. U.S. Geological
Survey Water- Resources Investigation Report 98-4005, 90 p.
Hill M.C., Banta E.R., Harbaugh A.W. and Anderman E.R. (2000) MODFLOW-2000, the
U.S. Geological Survey modular ground-water model - User guide to the Observation,
Sensitivity, and Parameter-Estimation Processes and three post-processing programs:
U.S. Geological Survey Open-File Report 00-184, 210 p.
Hinton J.C. (1996) GIS and remote sensing integration for environmental applications.
International Journal of Geographical Information Systems, 10(7), pp.877–890.
72 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Holman I.P., Palmer R.C., Bellamy P.H. and Hollis J.M. (2005) Validation of an intrinsic
groundwater pollution vulnerability methodology using a national nitrate database.
Hydrogeol J, 13(5–6), pp.665–674.
Hrkal Z. (2001) Vulnerability of groundwater to acid deposition, Jizerske Mountains,
northern Czech Republic: construction and reliability of a GIS-based vulnerability map,
Hydrogeol J., 9, pp.348–357.
Jackson T.J. (2002) Remote sensing of soil moisture: implications for groundwater recharge,
Hydrogeology Journal, 10, pp.40–51.
Jain A.K. (1989) Fundamentals of image processing: Prentice Hall, Englewood Cliffs, New
Jersey, p.342–357.
Jensen J. (1996) Introductory digital image processing: Prentice Hall Series in Geographic
Information Science, Englewood Cliffs, New Jersey, pp.159–165.
Jha M.K., Chowdhury A., Chowdary V.M. and Peiffer S. (2007) Groundwater management
and development by integrated remote sensing and geographic information systems:
prospects and constraints, Water Resour. Manage., 21, pp.427–467, DOI
10.1007/s11269-006-9024-4.
Johansson P.O., Scharp C., Alveteg T. and Choza A. (1999) Framework for ground-water
protection the Managua Ground Water System as an example, Ground Water, 37(2),
pp.204–213.
Kim Y.J. and Hamm S. (1999) Assessment of the potential for ground water contamination
using the DRASTIC/ EGIS technique, Cheongju area, South Korea. Hydrogeol J., 7(2),
pp.227– 235.
Kinzelbach W. (1986) Groundwater Modeling: An Introduction with Sample Programs in
BASIC. Elsevier, New York, 333 p.
Klimchouk A.B., Ford D.C., Palmer A.N. and Dreybrodt W. (eds) (2000) Speleogenesis,
evolution of karst aquifers. National Speleological Society, Inc., Huntsville, Alabama,
527 pp.
Knox R.C., Sabatini D.A. and Canter L.W. (1993) Subsurface Transport and Fate Processes.
Lewis Publishers, Boca Raton, Florida, 430 p.
Konikow L.F. and Grove D.B. (1977) Derivation of Equations Describing Solute Transport
and Dispersion in Ground Water, U.S. Geological Survey Water-Resources Investigations
77-19, 30 p.
Konikow L.F. (1978) Calibration of Ground-Water Models, in Verification of Mathematical
and Physical Models in Hydraulic Engineering, American Society of Civil Engineers,
New York, p.87-93.
Kouli M., Alexakis D., Vallianatos F., Soupios P., Sarris A., Xepapadaki A. and
Zoumpouloglou N. (2005) Integration of Geographic Information Systems in
Technological Education. An example in the Geo-environmental sciences, Proceedings
of the WSEAS 2005, Engineering Education, 12-14 July, Vouliagmeni, Athens, ISBN
960-8457-28-9, pp. 487-492.
Kouli M., Soupios P. and Vallianatos F. (2007a) GIS – based water management in the
Chania area, Western Crete, Studies in Mechanics, Environment and Geoscience, (eds.) J.
Krope, S. Necasova, N. Tutyshkin, E. Sapountzakis and M. Mucciarelli, WSEAS Press,
ISBN: 978-960-8457-70-6.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 73

Kouli M., Vallianatos F., Soupios P. and Alexakis D. (2007b) GIS - based morphometric
analysis of Keritis and Tavronitis drainage networks, Western Crete, Greece, Journal of
Environmental Hydrology, 15, pp. 1-17.
Lake I.R., Lovett A.A. and Hiscock K.M. (2003) Evaluating factors influencing groundwater
vulnerability to nitrate pollution: developing the potential of GIS. J Environ. Manage.,
68, pp.315– 328.
Lasserre F., Razack M. and Banton O. (1999) A GIS-linked model for the assessment of
nitrate contamination in groundwater, Journal of Hydrology, 224(3-4), pp. 81-90.
Lillesand T.M. and Kiefer R.W. (2000) Remote Sensing and Image Interpretation. Fourth
Edition, John Wiley & Sons, Inc., New York, pp. 724.
Lin Yu-F., Walker D. and Meyer S. (2003) Groundwater Flow Models of Northeastern
Illinois a case study for building MODFLOW models with GIS, Department of Natural
Resources, Technical Report.
Lo C.P. and Yeung A.K.W. (2003) Concepts and Techniques of Geographic Information
Systems. Prentice-Hall of India Pvt. Ltd., New Delhi, pp. 492.
Loague K., Bernknopf R.L., Green R.E. and Giambelluca T.W. (1996) Uncertainty of
groundwater assessments for agricultural regions in Hawaii, Review. J. Environ. Qual.,
25, pp.475–490.
Loague K. and Corwin D.L. (1998) Regional-scale assessment of non-point source
groundwater contamination, Hydrological Processes, 12(6), pp.957–966.
Lobo Ferreira J.P. and Oliveira M.M. (1997) DRASTIC groundwater vulnerability mapping
of Portugal. In: Groundwater: An Endangered Resource. Proceedings of Theme C of the
27th Congress of the International Association for Hydraulic Research, San Francisco,
USA.
Longley P.A., Brooks S.M., McDonnell R. and Macmillan B. (eds) (1998) Geocomputation:
A Primer, John Wiley & Sons Ltd., Chichester, pp. 290.
Longo A., Andreo B., Carrasco F., Cucchi F., Vìas J.M. and Jimènez P. (2001) Comparison
of two contamination vulnerability maps obtained by the SINTACS method in two
carbonate aquifers (S Spain). In: Proceedings of the 7th Conference on Limestone
Hydrology and Fissured Media. Besançon, Francia, pp 233–236.
Lynch S.D., Reynders A.G. and Schulze R.E. (1997) A DRASTIC approach to ground water
vulnerability in South Africa, S African J Sci., 93(2), pp.59–60.
Margat J. (1968) Vulnerabilite des nappes d’eau souterraine a la pollution [Ground water
vulnerability to contamination]. Bases de la cartographic, (Doc.) BRGM, 68 SGL 198
HYD, Orleans, France [in French]
McDonald M.G. and Harbaugh A.W. (1988) A Modular Three-Dimensional Finite Difference
Ground-Water Flow Model, USGS TWRI Chapter 6-A1, 586 p.
Meijerink A.M.J. (2000) Groundwater. In: Schultz GA, Engman ET (eds), Remote Sensing in
Hydrology and Water Management. Springer, Berlin, pp.305–325
Melloul M. and Collin M. (1998) A proposed index for aquifer water quality assessment: the
case of Israel’s Sharon region. J Environ Manage, 54(2), pp.131–142.
Merchant J. (1994) GIS-Based groundwater pollution hazard assessment: a critical review of
the DRASTIC model, Photogram Eng Remote Sensing, 60(9), pp.1117–1127.
Merritt L.M. and Konikow L.F. (2000) Documentation of a computer program to simulate
lake-aquifer interaction using the MODFLOW ground-water flow model and the
74 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

MOC3D solute-transport model: U.S. Geological Survey Water-Resources Investigations


Report 00-4167, 146 p.
Musa K.A., Akhir J.M. and Abdullah I. (2000) Groundwater prediction potential zone in
Langat Basin using the integration of remote sensing and GIS. www.GISdevelopment.net
(accessed on July 24, 2003)
Myers V.I. and Moore D.G. (1972) Remote sensing for defining aquifers in glacial drift.
Proceedings of the 8th International Symposium on Remote Sensing of the Environment,
Environmental Research Institute of Michigan, Ann Arbor, MI, pp. 715–728.
National Research Council (1993) Ground water vulnerability assessment: predicting relative
contamination potential under conditions of uncertainty. Committee on techniques for
Assessing Ground Water Vulnerability. Water Science Technology Board.
Nefedov K.E. and Popova T.A. (1972) Deciphering of groundwater from aerial photographs.
Amerind, New Delhi.
NRA (1994) Policy and practice for the protection of groundwater. Bristol: UK National
Rivers Authority.
Orzol L.L. (1997) User's guide for MODTOOLS: Computer programs for translating data of
MODFLOW and MODPATH into geographic information system files: U.S. Geological
Survey Open-File Report 97-240, 86 p.
Panagopoulos G.P., Antonakos A.K. and Lambrakis N.J. (2006) Optimization of the
DRASTIC method for groundwater vulnerability assessment via the use of simple
statistical methods and GIS, Hydrogeology Journal, 14, pp.894–911.
Perrin J., Pochon A., Jeannin P.Y. and Zwahlen F. (2004) Vulnerability assessment in karstic
areas: validation by field experiments, Environ. Geol., 46, pp.237–245.
Pinder G.F. and Bredehoeft J.D. (1968) Application of the Digital Computer for Aquifer
Evaluation, Water Resources Research, 4, pp.1069-1093.
Pinder G.F. (2002) Groundwater Modeling Using Geographical Information Systems.
JohnWiley&Sons, New York, pp. 248.
Piscopo G. (2001) Lachlan catchment groundwater vulnerability map explanatory notes,
Department of Land and Water Conservation, Sydney, New South Wales.
Pollack D.W. (1988) Semianalytical Computation of Path Lines for Finite Difference Models.
Ground Water, 26(6), pp.743-750.
Pollack D.W. (1989) Documentation of Computer Programs to Compute and Display
Pathlines using Results from the U.S. Geological Survey Modular Three-Dimensional
Finite-Difference Ground-Water Flow Model, USGS Open File Report 89-391, 188 p.
Praharaj T., Swain S.P., Powell M.A., Hart B.R. and Tripathy S. (2002) Delineation of
groundwater contamination around an ash pond Geochemical and GIS approach. Environ
Int., 27, pp.631–638.
Pratt W.K. (1991) Digital image processing: John Wiley & Sons, New York, 698 p.
Price C.V. and Pierce R.R. (1994) GIS Applications to hydrologic modeling in the U.S.
Geological Survey, Water Resources Division: Past and Future, in Proceedings of the
Fourteenth Annual ESRI User Conference, p.1163-167.
Rao P.S.C. and Alley W.M. (1993) Pesticides. In: Alley WM (ed) Regional groundwater
quality. Van Nostrand Reinhold, New York, pp.345–382.
Ray I.A. and Odell P.W. (1993) DIVERSITY: a new method for evaluating sensitivity of
groundwater to contamination, Environ. Geol., 22, pp.344–352.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 75

Reilly T.E., Franke O.L., Buxton H.T. and Bennett G.D. (1987) A Conceptual Framework for
Ground-Water Solute-Transport Studies with Emphasis on Physical Mechanisms of
Solute Movement. U.S. Geological Survey Water-Resources Investigation Report 87-
4191, 44 p.
Revzon A.L., Burleshin M.I., Krapilskaya N.M., Sadov A.V., Svitneva T.V. and Semina N.S.
(1983) Study of the desert geological environment with the aid of aerial and space
imagery. All-Union Sci. Res. Inst. Hydrol. Eng. Geol., Moscow.
Rhind D. (1989) Why GIS? ARC News, Vol. 11, No. 3, ESRI, Inc., Redlands, CA
Rigol J.P. and Chica-Olmo M. (1998) Merging remote sensing images for geological—
environmental mapping: application to the Cabo de Gata—Nijar Natural Park, Spain:
Environmental Geology, 34(2–3), pp.194–202.
Robins N., Adams B., Foster S. and Palmer R. (1994) Groundwater vulnerability mapping:
the British perspective. Hydrogéologie 3, pp.35–42.
Rokos D., Argialas D., Mavratza R., Seymour-St. C., Vamvoukakis K., Kouli M., Lamera S.,
Paraskevas H., Karfakis I. and Denes G. (2000) Structural Analysis for Gold
Mineralization Using Remote Sensing and Geochemical Techniques in a GIS
Environment: Island of Lesvos, Hellas, Nat. Resour. Res., 9, pp.277–293.
Ross M.A. and Tara P.D. (1993) Integrated hydrologic modeling with geographic information
systems, Journal of Water Resources Planning and Management, ASCE 119(2), pp.129–
141.
Rundquist D.C., Peters A.J., Di L., Rodekohr D.A., Ehrman R.L. and Murray G. (1991)
Statewide groundwater-vulnerability assessment in Nebraska using the DRASTIC/ GIS
model, Geocarto Int., 6(2), pp.51–57.
Rupert M.G. (2001) Calibration of the DRASTIC ground water vulnerability mapping
method, Ground Water, 39(4), pp.625–630.
Russ J.C. (1992) The image processing handbook: CRC Press, Boca Raton, Florida, 445 p.
Schultz G.A. (1993) Application of GIS and remote sensing in hydrology. In: Kovar K,
Nachtnebel HP (eds), Application of Geographic Information Systems in Hydrology
andWater Resources Management, IAHS Pub.No.211, pp.127–140.
Secunda S., Collin M.L. and Melloul A.J. (1998) Groundwater vulnerability assessment using
a composite model combining DRASTIC with extensive agricultural land use in Israel’s
Sharon region, J. Environ. Managept., 54, pp.39–57.
Shafer J.M. (1987) Reverse Pathline Calculation of Time-Related Capture Zones in
Nonuniform Flow, Ground Water, 25(3), pp.283-289.
Shukla S., Mostaghimi S., Shanholt V.O., Collins M.C. and Ross B.B. (2000) A county-level
assessment of ground water contamination by pesticides, Ground Water Monitor Rev.,
20(1), pp.104– 119.
Simonett D.S. (1983) The development and principles of remote sensing. In: Colwell RN
(ed), Manual of Remote Sensing. The American Society of Photogrammetry, VA pp. 1–
35
Sinreich S., Kozel R. and Mudry J. (2004) Specific vulnerability method. In: Zwahlen F (ed)
Vulnerability and risk mapping for the protection of carbonate (karst) aquifers, final
report COST Action 620 European Commission Directorate-General for Research, EUR
20912, pp 67–83.
76 Maria Kouli, Nikos Lydakis-Simantiris and Pantelis Soupios

Soupios P. and Stavroulakis G. (2006) Groundwater Flow Modeling in Keritis Basin,


Technical Report in the Framework of the project: Multiparametric analysis in karstic
river basins: Keritis river Basin (in greek).
Soupios P., Kouli M., Vallianatos F., Vafidis A. and Stavroulakis G. (2007) Estimation of
Aquifer Parameters from Surficial Geophysical Methods: A Case Study of Keritis Basin
in Crete, Journal of Hydrology, 338, pp.122-131, doi: 10.1016/j.jhydrol.2007.02.028.
Srinivasan R. and Arnold J.G. (1994) Integration of a Basin Scale water quality model with
GIS, Journal of the American Water Resources Association, 30(3), 453–462
doi:10.1111/j.1752-1688.1994.tb03304.x
Stafford DB (ed) (1991) Civil engineering applications of remote sensing and geographic
information systems. ASCE, New York.
Tesoriero A.J., Inkpen E.L. and Voss F.D. (1998) Assessing ground-water vulnerability using
logistic regression. Proceedings for the Source Water Assessment and Protection 98
Conference, Dallas, TX, pp.157–165.
Thapinta A. and Hudak P.F. (2003) Use of geographic information systems for assessing
groundwater pollution potential by pesticides in Central Thailand, Environ. Int., 29(1),
pp.87–93.
Todd D.K. (1980) Groundwater hydrology. 2nd edition, John Wiley & Sons NY, pp. 111–
163.
Torak L.J. (1992a) A MODular Finite-Element model (MODFE) for areal and axisymmetric
ground-water-flow problems, part 1: model description and user's manual: U.S.
Geological Survey Open-File Report 90-194, 153 p.
Torak L.J. (1992b) A MODular Finite-Element model (MODFE) for areal and axisymmetric
ground-water-flow problems, part 3: design philosophy and programming details: U.S.
Geological Survey Open-File Report 91-471, 261 p.
Toth J. (1963) A Theoretical Analysis of Groundwater Flow in Small Drainage Basins,
Journal of Geophysical Research, 68, pp.4795-4812.
Tsagarakis P.K., Dialynas E.G. and Angelakis N.A. (2004) Water resources management in
Crete (Greece) including water recycling and reuse and proposed quality criteria,
Agricultural Water Management, 66(1), pp.35-47.
Tsihrintzis V.A., Hamid R. and Fuentes H.R. (1996) Use of geographic information systems
(GIS) in water resources: a review, Water Resources Management, 10, pp.251–277.
U. S. Environmental Protection Agency (1996) Environmental indicators of water quality in
the United States: Washington, D.C., Office of Water, EPA 841-R-96-002, EPA,
Washington, DC
USGS (1997) Geographic Information Systems. An Information Brochure, USGS, Reston,
VA
Van de Griend A.A., Camillo P.J. and Gurney R.J. (1985) Discrimination of soil physical
parameters, thermal inertia and soil moisture from diurnal surface temperature
fluctuations, Water Resources Research, 21, pp.997–1009.
Van der Laan F. (1992) Raster GIS allows agricultural suitability modeling at a continental
scale, GIS World, pp. 42–50.
Van Stempvoort D., Ewert L. and Wassenaar L. (1993) Aquifer vulnerability index (AVI): A
GIS compatible method for groundwater vulnerability mapping, Can. Water Res. J., 18,
pp.25–37.
GIS-Based Aquifer Modeling and Planning Using Integrated Geoenvironmental… 77

Vías J.M., Andreo B., Perles M.J. and Carrasco F. (2005) A comparative study of four
schemes for groundwater vulnerability mapping in a diffuse flow carbonate aquifer under
Mediterranean climatic conditions. Environ Geol 47(4), pp.586–595.
Vías J.M., Andreo B., Perles M.J., Carrasco F., Vadillo I. and Jimenez P. (2006) Proposed
method for groundwater vulnerability mapping in carbonate (karstic) aquifers: the COP
method - Application in two pilot sites in Southern Spain, Hydrogeology Journal, 14, pp.
912–925, DOI 10.1007/s10040-006-0023-6
Vrba J. and Zaporozec A. (1994) Guidebook on mapping groundwater vulnerability.
International Association of Hydrogeologists. International contributions to
hydrogeology, 16, Heise, Hannover.
Wang H.F. and Anderson M.P. (1982) Introduction to Groundwater Modeling. W.H.
Freeman and Company, San Francisco, CA, 237 p.
Wang Y., Merkel B.J., Li Y., Ye H., Fu S. and Ihm D. (2007) Vulnerability of groundwater in
Quaternary aquifers to organic contaminants: a case study in Wuhan City, China,
Environmental Geology, 53(3), pp.479-484.
Waters P., Greenbaum P., Smart L. and Osmaston H. (1990) Applications of remote sensing
to groundwater hydrology, Remote Sensing Review, 4, pp.223–264.
Watkins D.W., McKinney D.C., Maidment D.R. and Lin M. (1996) Use of Geographic
Information Systems in Ground-Water Flow Modeling, Journal of Water Resources
Planning and Management, 122(2), pp.88-96, doi 10.1061/(ASCE)0733-9496 (1996)
122: 2(88)
WHO, World Health Organization (2004) Guidelines for drinking-water quality, vol 1, 3rd ed,
recommendations. WHO, Geneva, Switzerland, pp 145–220
Zabet T.A. (2002) Evaluation of aquifer vulnerability to contamination potential using the
DRASTIC method, Environ. Geol., 43, pp.203–208.
Zhang H., Haan C.T. and Nofziger D.L. (1990) Hydrologic modeling with GIS: An overview.
Applied Engineering in Agriculture, ASAE, 6(4), pp.453–458.
Zheng C. (1990) MT3D: A Modular Three-Dimensional Transport Model for Simulation of
Advection, Dispersion, and Chemical Reactions of Contaminants in Ground-Water
Systems, U.S. EPA, R.S. Kerr Environmental Research Laboratory, Ada, Oklahoma,
170 p.
Zheng C. (1991) PATH3D, A Ground-Water Path and Travel-Time Simulator, User's
Manual, S.S. Papadopulos & Associates, Inc. Bethesda, MD, 50 p.
Zheng C. and Bennett G.D. (1995) Applied Contaminant Transport Modeling. Van Nostrand
Reinhold, New York, 440 p.
Zheng C., Hill M.C. and Hsieh P.A. (2001) MODFLOW-2000, the U.S. Geological Survey
modular ground-water model - User guide to the LMT6 package, the linkage with
MT3DMS for multi- species mass transport modeling: U.S. Geological Survey Open-File
Report 01-82, 43 p.
Zwahlen F. (ed) (2004) Vulnerability and risk mapping for the protection of carbonate (karst)
aquifers. Final Report (COST action 620), European Commission, Brussels, 315.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 79-111 © 2008 Nova Science Publishers, Inc.

Chapter 2

A MODELING APPROACH SUPPORTING AIR


SPARGING SYSTEM DESIGN

Sabrina Saponaro, Sara Puricelli, Elena Sezenna


and Luca Bonomo
Politecnico di Milano, Department of Environmental, Hydraulic, Infrastructures and
Surveying Engineering - Environmental Section, Milano, Italy 20133

Abstract
Air Sparging (AS) is a remediation technique for groundwater based on the injection of
pressurized air under the water table to strip the volatile organic pollutants from the saturated
zone; moving in soil towards the ground surface, pollutants are captured by a Soil Vapor
Extraction (SVE) system. The effectiveness of the treatment depends on pollutant volatility,
and air flow behavior in soil, that is affected by soil properties (intrinsic permeability,
heterogeneity, anisotropy, etc.), plant configuration (number, location, depth and screen
position of AS injection and SVE extraction wells) and values of the process parameters (AS
and SVE flowrates, AS injection cycles, etc.). Airflow in soil can be described by the
multiphase flow theory, which is based on a closed system of partial differential equations,
whose solution results in air pressure and air saturation in soil as a function of space and time;
the numerical finite difference code TOUGH2 (v. 2.0) can be used as a tool to solve the
system of equations. This paper shows the application of this tool to assess the effects of
different AS injection and SVE extraction flow rates, sparging mode, surface sealing, number,
location and screen position of AS or SVE wells for the remediation of a polluted site in the
northwest of Milan (Italy).

Technology Principles and Application Field


Air Sparging (AS) is a remediation technique for unconfined aquifers, based on the injection
of pressurized air under the water table to strip the volatile organic pollutants from the
saturated zone; moving in soil towards the ground surface, pollutants are captured by a Soil
Vapor Extraction (SVE) system. An AS treatment differs from a Biosparging (BS) treatment,
as during BS the injected air is aimed at delivering oxygen in groundwater in order to
80 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

promote the aerobic biodegradation of pollutants, eventually limited by anaerobic conditions,


under the groundwater level; BS will not be taken into account in this chapter.
The feasibility and effectiveness of an AS treatment strictly depends on the following
factors:

a) soil intrinsic permeability under the water table, which must be higher than about 10-9
m2. In order to allow vapor extraction by the SVE system, soil intrinsic permeability
in the vadose zone must be higher than 10-14 m2;
b) vapor/dissolved phase partitioning of the pollutants, affected by the Henry’s law
constant (H), whose limiting value for the application of AS is usually assumed to be
1 Pa m-3 mol-1 at 20 °C; high values of H promote pollutant partitioning towards the
gaseous phase, which is the phase directly affected by the injected air flow.

Empirical System Design and Operation


The design of the treatment system should be oriented to optimize the following parameters:

• depth and screen position, for both the AS and SVE wells;
• AS injection and SVE flow rates;
• number of wells necessary to cover the entire contaminated zone, optimal location
and operation, for both the injection system and the extraction system.
• Except for well depth and screen position, the design of an AS plant is essentially
based on results of pilot tests, which should be performed to assess:
• the characteristic curve of the well, that is the air flow rate injected vs the injection
pressure applied at the well head;
• the duration of the transient step following the injection system shut-on/shut-off;
• the amplitude of the Zone Of Treatment (ZOTAS), that is where an air saturation over
10% can be achieved, and the air flow is assumed to be effective in stripping the
pollutants (McCray et al., 1996; Suthersan, 1997).

ZOTAS can be estimated under different operating conditions by using neutron probes, or
electrical resistivity tomography, which result in the variation of 2- or 3-D distribution of
water content in soil compared to the distribution before air injection has started up.
Unfortunately, these techniques are expensive and the findings often difficult to interpret, so
that the measurement of ZOTAS is generally replaced by the measurement of the Zone Of
Influence (ZOIAS), where a variation in the piezometric level can be observed with respect to
the natural condition, for example by pressure transducers. Even if ZOIAS is easier to measure
than ZOTAS, it overestimates the amplitude of subsoil effectively treated (Lundegrad, 1995;
USACE, 2008; USEPA, 2001).
Pilot tests also involve the SVE system, in order to acquire data for the proper design of
the off-gas treatment section, and to estimate the amplitude of the Zone Of Influence
(ZOISVE). ZOISVE is defined as the maximum distance from the extraction well where a
perturbation in air pressure (ΔPg) can be observed with respect to natural conditions.
Actually, a more reliable amplitude of the zone where the SVE system is effective in
A Modeling Approach Supporting Air Sparging System Design 81

removing pollutants is the Zone Of Treatment (ZOTSVE), where ΔPg exceeds 25 Pa (Dixon et
al., 2006) or, alternatively, air velocity is higher than 1.0 10-4 m s-1 (USACE, 2002).

Pilot Plant

An AS+SVE pilot plant (Fig. 1) is usually composed by:

• one injection well (PAS) and one extraction well (PSVE), equipped with monitoring
and control devices at the well head (pressure gauge, air flow meter, pressure relief
valve, and flow control valve). Vertical wells are usually used, 25 mm to 100 mm in
diameter; treatment effectiveness does not seem significantly affected by well
diameter, but high frictional losses can result in small deep wells. Screen length of
PAS ranges between 0.3 m and 0.7 m, because air often escapes within a very short
interval near the top of longer well screens; the top of the screen is located at about
1.5 m below pollution (USACE, 2008; USEPA, 2004);
• an air compressor or a blower connected to PAS, and a blower or a vacuum pump
connected to PSVE;
• 3 to 9 monitoring wells for groundwater and the vadose zone, located in various
directions (to assess soil heterogeneities) and at various distances from PAS and PSVE,
according to the expected ZOIAS and ZOISVE. Monitoring wells for groundwater are
equipped for the measurement of the piezometric level, whereas the air pressure is
measured in the vadose zone.

Figure 1. AS + SVE pilot plant.


82 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Pilot Test

The first portion of the pilot test should be conducted using PSVE only. A vacuum is applied at
the extraction well, and resulting airflow rate QV,SVE (m3 s-1), soil gas vacuum levels, and
effluent contaminant concentrations are measured (USACE, 2002).
The second portion of the test involves air injection through PAS. Injection pressure Pi
(Pa) and airflow QV,AS (m3 s-1) are monitored at PAS wellhead, as well as pressure below water
table and in the vadose zone (USACE, 2008). In order to get airflow in the saturated zone, the
injection pressure should have a minimum value Pmin (Pa), which takes into account frictional
losses in piping and well, the filter pack air-entry pressure, the formation air-entry pressure,
and the hydrostatic pressure. However, frictional losses in piping and well can be often
neglected. The filter pack air-entry pressure tends to be quite small (i.e., < 1 kPa for uniform
sands commonly used as filter pack). Air-entry pressure of formations ranges from negligible
values (below 1 kPa) for coarse-textured media such as coarse sands and gravels, to about few
hundreds of kPa for clays. At most sites, the key component of the injection pressure is the
hydrostatic pressure Ph (Pa) needed to displace the column of water in the well pipe before
starting the injection:

Ph = ρl g z wc (1)

where ρl (kg m–3) is the water density, g is the gravitational acceleration (m s-2), and zwc (m) is
the difference between depth to the top of PAS well screen, and pre-sparging depth to the free-
water surface within the sparging well; therefore, it can be usually assumed Pmin = Ph.
The injection pressure Pi should be limited to a value that does not create excessive
overpressurization, so that aquifer fracturing does not occur; a breakthrough of the aquifer
will result in short-circuiting of the airflow, greatly reducing the effectiveness of the
treatment. As a general guideline, the overburden pressure is estimated as the pressure exerted
by the weight of the soil and water column overlying PAS well screen:

Pf = ρs (1 - f) g z tws + ρl g f z wc (2)

where ρs is soil mineral density (kg m-3), f is soil porosity (-), and ztws is the depth of the well
screen top from ground surface. The maximum injection pressure Pmax (Pa) is usually
calculated as:

Pmax = fs Pf (3)

where fs (-) is a design safety factor (0.6 to 0.8) (USACE, 2008).

Full-Scale Plant Operation

Full-scale plants are usually equipped as pilot plants, but more injection and extraction wells
are typically installed; the number of SVE wells is usually lower than for AS, because for a
specific soil extending both in the vadose zone and in the water-saturated zone, ZOISVE is
A Modeling Approach Supporting Air Sparging System Design 83

larger than ZOIAS. Air sparging wells should be placed so that the overlapping of their ZOTAS
completely covers the area of contamination (USEPA, 2004). The depth to the top of the
screened interval of AS wells may differ according to the pollution distribution in the aquifer.
Typical values for the volumetric injection flowrates through AS wells QV,AS range from 5
Nm h to 25 Nm3 h-1. Twice to four times values are suggested for ⏐QV,SVE⏐ (NFESC, 2001;
3 -1

USACE, 2008).

Figure 2. Behavior of air flow in a homogeneous medium (modified after Lundegard et al., 1995).

AS treatment may be operated by continuous or pulsed air injection (USACE, 2008).


Actually, when pressurized air is injected through PAS into the saturated medium, it dewaters
the formation creating a desaturated zone, with a characteristic reversed cone shape. Three
temporal phases can be observed: expansion, collapse, and steady state (Lundegard, 1995;
Lundegard et al., 1995; Thomson et al., 2000). In the ideal case of a homogeneous medium,
the short period of “spherical” growth near the injection point [Fig. 2(a)] is followed by the
upward motion due to buoyancy [Fig. 2(b)]. The rate of air injection into groundwater
exceeds the rate of airflow flowing through the vadose zone, causing water mounding and air
migration under the water table. Once airflow increases in the vadose zone, the pressure
below the water table drops, so that the air zone contracts until equilibrium is reached
between air injection and air leakage, and dissipation of mounding occurs too [Fig. 2(c)]. The
duration of the transient state may vary between few minutes, for high permeability aquifers,
and a few days (University of Wisconsin, 2000). Air injection under the water table promotes
groundwater mixing, occurring at a larger extent during the transient state than at steady state;
therefore, pulsing may result in enhancing mass transfer phenomena between the injected air
and groundwater. Cycling from one sparging well to another using the same compressor is
expected to provide a cost savings because of smaller gas compressor requirements and
reduced energy costs (Marley et al., 1994).

Air and Water Flow in Porous Media


As previously said, the design and operation of AS systems has remained largely empirical.
One of the most significant factor affecting AS performance is the air distribution in the target
treatment zone; therefore, modeling tools able to describe the airflow path in soil can be
useful in refining the treatment.
84 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

General Concepts

A soil system is composed of three phases (β): solid, liquid (β = l) and gas (β = g). The
degree of saturation of the mobile phases, Sβ (−), is defined as the ratio between the volume of
phase β, Vβ, and the pore volume, Vf:


Sβ = (4)
Vf

The degree of saturation is related to the volumetric phase content through soil total
porosity f (-):

θβ = f Sβ (5)

The following constitutive relation applies:

Sl + Sg = 1 (6)

Specific symbols can be introduced for specific values of Sβ:

• Sls and Sgs are the maximum values which water and air saturation may have,
respectively;
• Slr is the residual liquid saturation, defined as the degree of water saturation at which
“water is immobile or water flow is negligible on the time scale of importance for the
evaluation of flow properties” (Byrnes, 2008; Cornelis et al., 2005; Moseley et al.,
1996). Under this condition, water films on solid phase are so thin that the liquid
phase doesn’t follow suction forces (Luckner et al., 1989);
• Sgr is the residual gas saturation, which is the value below which the gas phase is
discontinuous and it does not flow any more (Luckner et al., 1989; Moseley et al.,
1996).

Relationships occurring among saturations are shown in Fig. 3.

Figure 3. Distribution of two immiscible phases in a porous medium, expressed as water or gas
volumetric phase content.

Under equilibrium conditions along a vertical soil profile, there is a static distribution of
Sl. Water table, where water is at the atmospheric pressure, is the upper boundary of the
saturated zone, with liquid pressure Pl (Pa) increasing with depth due to the hydrostatic
A Modeling Approach Supporting Air Sparging System Design 85

pressure. Above the water table, the unsaturated zone has pores partially filled by water and
partially by air. The capillary fringe, whose thickness ranges from a few centimeters for
coarse sands to several meters for clays (Berkowitz et al., 2004), is a transition zone where Sl
≤1 and Pl is below the air pressure Pg due the interfacial force pushing water through pores
(Corey, 1986). The difference of pressure across the interface of air and water is called
capillary pressure Pc (Pa):

Pc = Pl − Pg (7)

Capillary pressure can be expressed in term of pressure head, h (m):

h = Pc /(g ρl) (8)

Suction Ψ (m) is also used and defined as:

Ψ = – h. (9)

Figure 4. Pressure head (h) as a function of the degree of water saturation (Sl) for a coarse soil (I) and a
fine soil (II). Point “A” provides the air-entry value (ha)

The close relation between Sl (or θl) and Pc (or h) is described by the soil-water
characteristic curve, or soil-water retention curve. Fig. 4 shows an example for a coarse soil
(I) and a fine soil (II). As Sl decreases towards Slr, |h| tends to very high values. Point “A” is
also shown, locating the critical value of |h| that has to be exceeded in order to have air
intrusion in the saturated soil; it is called air-entry value, entry pressure head or bubbling
pressure head, ha, and it is the pressure head corresponding to the highest water saturation at
which air is interconnected (degree of saturation to air equal to Sgr) (Moseley at al., 1996) or
86 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

where dh/dSl has the maximum value (van Genuchten et al., 1985). In a given soil, the
relationship between |h| and Sl varies, depending on whether the soil is wetting or drying; this
discrepancy is referred to as hysteresis.

Measurement of Soil-Water Characteristic Curves

Soil-water characteristic curves can be obtained by field or laboratory measurements.


For field measurements, tensiometers can be used. A tensiometer usually consists of a
saturated porous cup of ceramic material, connected through a tube or a rigid body tube to a
pressure sensor; the cup has an air-entry value larger in magnitude than the lowest pore water
pressure to be measured. When the tensiometer is placed in soil, water flows through the
porous cup to soil or soil water flows to the porous cup, until equilibrium has been reached.
The operating range of tensiometers is from 0 kPa pressure head, when the soil is fully
saturated, to about -86 kPa, when soil is dry (ASTM, 2004; ISO, 1995).
For laboratory measurements, various methods are available. Since soil structure has a
strong influence on water-retention properties, the measurements should be performed on
undisturbed soil samples, according to ISO (2003). All methods operate to acquire data along
soil drying curve. They differ in how they are performed (if changing water content or
suction), how water content is measured (gravimetrically or volumetrically), and the range of
suction where they are applicable; in order to cover the entire range of water content, more
than one method is applied, overlapping data at the linking points. For a specific suction
range, it is possible to choose between different alternatives according to sample
characteristics. For data at nearly-saturated condition (low suction), and to determine the air-
entry value, it is possible to apply:

a) hanging column method (0 kPa to 80 kPa) for coarse soils with a small amount of
fines (ASTM, 2002);
b) porous plate and burette apparatus (0 kPa to 20 kPa) for soils with a weak structure
and unconsolidated sands, because samples are poorly disturbed (ISO, 1998);
c) Wind’s evaporation method (0 kPa to 70-80 kPa) (ISO, 2004; Meadows et al., 2005);
d) sand, kaolin or ceramic suction tables (0 kPa to 50 kPa) for samples with a wide
range of textures and organic matter content (ISO, 1998);
e) centrifuge method (0 kPa to 120 kPa) for coarse soils (ASTM, 2002).
At intermediate water contents and suctions:
f) pressure chamber, also called Richards chamber (5 kPa to 1500 kPa), suitable for
fine soils retaining water more tightly (ASTM, 2002), but less for heterogeneous soil
horizons or those with a strong structural composition (ISO, 1998).
Far from saturation and next to the dry end of the curve:
g) instantaneous profile method, through chilled mirror hygrometer (1000 kPa to 100
MPa) (ASTM, 2002) or Time Domain Reflectometry (TDR) (Slawinski et al., 2004),
suggested for dry soils.

Except for method (g), soil specimen is completely saturated and placed in contact with a
water-saturated porous plate. The time required for wetting depends on soil water content and
texture, being about one or two days for sands and two weeks or more for clayey soils.
A Modeling Approach Supporting Air Sparging System Design 87

Figure 5. Schemes of most applied apparata to acquire soil water retention curves, according to:
hanging column method (a) (modified after ASTM, 2002), kaolin suction table method (b) (modified
after ISO, 1998), pressure chamber method (c) (modified after ASTM, 2002), TDR method (d)
(modified after Slawinski et al., 2004).

In the hanging column method [Fig. 5(a)], suction is applied by reducing the pore water
pressure while maintaining the pore gas pressure at the atmospheric condition. The suction
supply consists of two reservoirs, whose relative elevation induces vacuum that moves water.
Water flows from the specimen until the equilibrium has been reached. The magnitude of the
applied suction is measured with a manometer, and the expelled water is volumetrically
measured. Appling different suctions, several point of the soil water characteristic curve can
be obtained.
The porous plate and burette apparatus is based on the principle of method (a), but it is
suitable within a shorter suction range. The apparatus consists of a funnel containing a porous
plate where the soil sample is located. A negative pressure is applied by lowering a leveling
burette, connected to the bottom of the funnel. At the equilibrium condition, the volume of
water in the burette is recorded, providing the volume of water withdrawn from the soil
sample. A variation of this method is given by Winfield et al. (2002) (USGS, 2008), who
proposed a large-core controlled-volume method with a relatively large sample (10 cm in
diameter, 15 cm height), to preserve the original feature of soil and to reduce time required to
reach the equilibrium condition.
88 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Wind’s evaporation method is a very simple technique. The soil sample, whose size may
change according to the soil type (generally 8 cm height and 10-15 cm in diameter), is
saturated and allowed to dry by evaporation from the top surface; water content is obtained by
weight measurements, whereas pressure heads are read through tensiometers (at least four)
located at various depths in the sample. The accuracy of this method is quite good, even if a
certain overestimation of the pressure head can occur (Garnier et al., 2004).
Suction tables [Fig. 5(b)] are suitable in laboratories where regular equipment
maintenance programs are implemented. Several saturated small soil cores are placed in
contact with the surface of the table (made of sand, kaolin or ceramic material), so that they
lose pore water until their suction is equivalent to that of the suction table. Soil water content
is measured by weighing.
In the centrifuge method, the specimen is contained in a support chamber that is
subjected to a centrifugal force, with acceleration up to 1000 g (Khaleel et al., 1995)
increased by steps. Water displaced from soil at a given angular velocity is measured in a
calibrated cylinder.
In pressure chambers [Fig. 5(c)], the pore water pressure is maintained at the atmospheric
pressure, and the pore air pressure is raised to apply the suction. Compressed air is generally
provided by an air compressor or bottled gas, and it is monitored through Bourdon gages,
mercury and water manometers, or pressure transducers. Samples are placed on a porous plate
or membrane, which constitute a saturated interface between the pore water and the water in
the volume measuring system. This surface is hydrophilic and has an air entry pressure
greater than the maximum matric suction applied during the test. Water content is measured
volumetrically or gravimetrically by weighing the specimen after removal from the apparatus.
Sample size is usually smaller than for other methods, generally 5 cm in diameter and 0.5 cm
to 1 cm height.
In chilled mirror hygrometer method, the sample (at least 2 cm in diameter and 0.5 cm
height) is weighted at the beginning of the test and then it is allowed to dry to lower water
content by exposure to the atmosphere. The variation in water content is measured
gravimetrically, and the corresponding suction is indirectly provided by a chilled mirror
hygrometer. TDR test is based on the same principle [Fig. 5(d)]. TDR probes are used to
measure soil humidity; microtensiometers are inserted through the sample (5 cm to 10 cm in
height, 5.5 cm in diameter) to get soil water potential (Slawinski et al., 2004).
Experimental data describing unsaturated soil behavior give basic information necessary
to apply the multiphase flow theory reported below.

Multiphase Flow Theory

Airflow in soil can be described by the multiphase flow theory, which is based on a closed
system of equations, whose solution results in air pressure Pg and air saturation Sg in soil as a
function of space and time. The main hypothesis of this theory is that air and water flow as a
continuum in soil and that the relative permeability of each soil phase depends on its degree
of saturation in soil (EOLBNL, 1999).
Under laminar flow conditions for both fluids, the flow rates are given by Darcy’s law:
A Modeling Approach Supporting Air Sparging System Design 89

qβ = −
k rβ k
μβ
(∇ P
β − ρ β ge z ) (10)

where q (m s-1) is Darcy velocity, kr (-) is the relative hydraulic conductivity, and μ (Pa s) is
the dynamic viscosity for each phase β, k (m2) is the intrinsic permeability of soil (II rank
tensor) and ez is a unit vector in z direction of a Cartesian coordinate system with origin at the
bottom of the domain and increasing upwards.
Water density can be assumed as a constant value, whereas the gas phase can be
considered as an ideal gas, whose density and pressure are linked by the following linear
relation:

ρg
Pg = RT (11)
ωg

where ωg (g mol-1) is the air molecular weight, R (Pa m3 mol-1 K-1) is the universal gas
constant, and T (K) is the temperature. Air is assumed to be compressible, differently from the
simplifying hypothesis used in analytical models (Van Dijke et al., 1995) not accurately
describing the air flow (Mei et al., 2002).
Continuity equations, expressing mass conservation, can be written as:

∂ (fρ β S β ) ⎛ ρ β k rβ k ⎞
∂t
= ∇• ⎜
⎜ μ
(∇ Pβ − ρ β ge z )⎟⎟ + Ξ β (12)
⎝ β ⎠

where Ξβ (kg m-3 s-1) are sinks or sources.


Assuming air as the phase of reference, Pl and Pg are linked through Pc according to Eq.
(7). Under isothermal conditions, the equations are solved for the independent variables Pg
and Pl, calculating ρg by Eq. (11), Pc by Eq. (7), Sg by Eq. (6) and Sl by the soil water
characteristic curve. Alternatively, other couples of variables can be selected, such as Pg and
Sg.
The solution of the previous equations in a domain (Ω) depends on: i) the initial
conditions (i.c.), which define the values for the variables at time t = 0 in Ω; and ii) the
boundary conditions (b.c.), specifying the values for the variables along the domain boundary
(Dirichlet b.c.) or the air and water flowrates through Ω boundary (Neumann b.c.).

Numerical Code TOUGH2


The previous equations can be solved by using the numerical finite difference code TOUGH2
v. 2.0 (EOLBNL, 1999), specifically the EOS3 module, and PetraSym v. 4.0 β (Thunderhead
Engineering, KS - USA) as pre- and post- interactive processor.
TOUGH2 is a simulation program for multi-dimensional multi-phase fluid and heat flows
in porous and fractured media. It complies with the ANSI X3.9-1978 (FORTRAN77)
standard, and has been developed with a modular architecture.
90 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

TOUGH2 solves the basic mass- and energy- balance equations written in the general
form:

d κ κ κ
∫ M d V j = ∫ F • n d Γ j + ∫ r dV j
dt V j
(13)
Γj Vj

The equation is referred to an arbitrary subdomain Vj of the system under study, which is
bounded by the closed surface Γj. Mκ represents mass or energy per volume, with κ = 1, ...,
NK labeling the mass components (water, air, ...), and κ = NK + 1 the heat component. F
denotes mass or heat flux, n is a normal vector on surface element Γj (pointing inward into
Vj), and r denotes sources or sinks.
The total mass of component κ (1 to NK) is obtained by summing over the fluid phase β
(liquid, gas):

M κ = f Σ S β ρ β Χ βκ (14)
β

where Χ βκ is the mass fraction of κ in phase β. The heat accumulation term is:

M NK +1 = (1 − f ) ρs C sT + f Σ S β ρ β u β (15)
β

where CS is the specific heat of the rock, and uβ is specific internal energy in phase β.
In order to solve numerically the equations, the continuous space is discretized by using
the integral finite difference method. Time is discretized as a first-order finite difference; time
steps are automatically adjusted by the code during a simulation run, according to the
convergence rate of the iteration process.
The thermophysical properties of fluid mixture needed for assembling equations (13) are
provided by "equation of state" (EOS) modules, where the fluid phase conditions are
recognized from the numerical values of primary variables. The various EOS modules can
represent different fluid mixtures; EOS3 applies to air and water mixtures. Air is
approximated as an ideal gas. The capillary pressure function (Pc vs Sl) and the relative
permeability function (krl and krg vs Sl) of the porous medium have to be provided. As an
option, the user can choose for an isothermal two-phase flow problem, so that the heat
balance equation is not engaged.
PetraSym can assist TOUGH2 users to provide input data and to visualize the output for
the variables of concern by 2- or 3-D plots at various times.

Capillary Pressure Functions

THOUGH2 includes the capillary functions discussed in the following.

No capillary pressure
The medium may have no capillary pressure, which means Pc = 0 apart from the Sl value.
A Modeling Approach Supporting Air Sparging System Design 91

Linear function
In order to get a preliminary assess of the behavior of the air injected in soil when no
experimental data about the capillary pressure function are available, a linear relationship
between Pc and the degree of saturation to water Sl can be selected:


⎪Pcr for S l ≤ S lr

⎪ S − Sl
PC = ⎨Pcr ls for S lr < S l < S ls (16)
⎪ S ls − S lr

⎪0 for S l ≥ S l s

where Pcr (Pa) is the minimum value of the capillary pressure may have.

Function of Pickens et al.


This is a nonlinear model based on the empirical equation proposed by King (Pickens et
al., 1979; van Genuchten et al., 1985) relating θl to h:

⎡ ⎧ ⎛ h ⎞
χ
⎫⎪ θ − θ ⎤
⎢ cosh⎪⎨ ⎜⎜ ⎟⎟ + ε ⎬ − l 0 lr cosh ε ⎥
⎢ ⎪⎩ ⎝ h0 ⎠ ⎪⎭ θ l 0 + θ lr ⎥
θl = θl0 ⎢ ⎥ (17)
⎢ ⎧⎪ ⎛ h⎞
χ

⎪ θ −θ ⎥
⎢ cosh⎨ ⎜⎜ ⎟⎟ + ε ⎬ + l 0 lr cosh ε ⎥

⎣ ⎪⎩ ⎝ h0 ⎠ ⎪⎭ θ l 0 + θ lr ⎥

where θl0, h0, χ and ε are curve-fitting parameters of a set of experimental values. This
expression showed good results in fitting experimental data for several coarse- and fine-
textured soils. Note that θl0 and h0 may refer to the saturated condition.
In case of ε = 0, the inversion of Eq. (17) results in:

1/ χ
⎧ ⎡ ⎛ 2 ⎞⎤ ⎫
⎪ ⎢ 1 + Sl ⎜ ⎛ S ⎞ ⎟⎥ ⎪
⎪⎪ ⎢ Sl 0 Sl 0 − Slr ⎜ ⎜ 1+ l ⎟ ⎟⎥ ⎪⎪
l 0 Sl 0 − Slr
⎜1 + 1− ⎜ ⎟
S
Pc = P0 ⎨1− ⎢ ⎟⎥ ⎬ (18)
⎜ ⎟
⎪ ⎢ 1− Sl Sl 0 + Slr ⎜
⎜ 1−
Sl Sl 0 + Slr

⎟⎥ ⎪
⎪ ⎢ Sl 0 ⎜⎜ ⎝ S ⎠ ⎟⎟⎥ ⎪
⎪⎩ ⎣ ⎠⎦ ⎪⎭
l0

where P0 is related to h0 by Eq. (8), and Sl0 and Slr are related to θl0 and θlr respectively,
through porosity.

Milly’s function
This function is based on the generalization of the soil water retention curve obtained by
Haverkamp for the Yolo clay soil, as reported in Milly (1982):
92 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

⎧ ⎡ 4

−1

⎪0.371⋅ ⎢1 + ⎛⎜ log( −h ) ⎞⎟ ⎥ + 0.124 for h < −0.01 m


⎪⎪ ⎢⎣ ⎝ 2.26 ⎠ ⎥⎦
θl = ⎨ (19)

⎪0.495 for h ≥ −0.01 m
⎪⎩

Capillary pressure is calculated according to the following expression:

⎧ ⎡ 0.371 ⎤ 4
1

⎪ 2 . 26 ⎢
S −S
−1⎥
⎪− 97.783 ⋅ 10 ⎣ l lr ⎦ for S l − S lr > 0.371
Pc = ⎨ (20)

⎪− 97.783 for S l − S lr ≤ 0.371

where the coefficient 97.783 is the factor to convert 0.01 m head to a pressure value in Pascal.
Within Eq. (20), θl has been replaced by Sl, without changing the numerical coefficient 0.371,
therefore neglecting total porosity. Furthermore, Slr in Eq. (20) is an input data, replacing the
coefficient 0.124 in Eq. (19), which was specific for the Yolo clay soil.

Leverett’s function
The semi-empirical Leverett’s function, which is suitable for unconsolidated sands, was
rearranged by Udell et al. (1985), resulting in a polynomial dependence of Pc on the effective
liquid saturation Sel:

[
Pc = P0 σ 1.417 ⋅ (1 − Sel ) − 2.120 ⋅ (1 − Sel ) 2 + 1.263 ⋅ (1 − Sel ) 3 ] (21)

where σ is the surface tension of water (supplied internally in TOUGH2, according to the
temperature), and

S l − S lr
Sel = (22)
S ls − S lr

where Sls is set equal to 1.

van Genuchten function


This function is based on the van Genuchten (1980) model, relating the dimensionless
water content θel:

θ l − θ lr
θ el = (23)
θ ls − θ lr

to h according to the following expression:


A Modeling Approach Supporting Air Sparging System Design 93

m
⎡ 1 ⎤
θ el = ⎢ n ⎥
(24)
⎣1 + (α h ) ⎦

where α, and n are positive empirical coefficients, and:

1
m = 1− (25)
n

Generally, α ranges from 0.5 m-1 to 5 m-1 for very fine to coarse soil, and m ranges from
0.3 to 0.6 depending irregularly on the percentage of sand and silt in soil (Mei et al., 2002).
Taking into account that θel = Sel, Eq. (24) can be inverted, resulting in the following
expression:

h=
1
α
(
(Sel )− 1m − 1 )
(1− m )
(26)

In TOUGH2, the capillary function is calculated as:

(
Pc = P0 (Sel )
−1
m −1 ) (1− m )
(27)

where:

ρl g
P0 = − (28)
α

Relative Permeability Functions

Two different groups of models for the relative permeability function are dealt in TOUGH2.
The first one includes mathematical expressions that do not have physical meaning but they
are often used in numerical codes (all phases perfectly mobile, and linear functions). The
second group includes functions relating krl to the degree of saturation to water according to
analytical models reported in literature and fitting experimental results (Corey’s curve,
Grant’s curves, Fatt and Klikoff model, van Genuchten-Mualem model, Verma et al. model).

All phases perfectly mobile


In this model both phases are perfectly mobile in the medium and no influence between
the two is taken into account. Therefore:

k rl = 1 for all saturations (29)


krg = 1 for all saturations (30)

Linear functions
The following functions are used by selecting this option:
94 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

1 Slr
k rl = Sl − (31)
Sls − Slr Sls − Slr

Sgr
k rg =
1
(1 − Sl ) − (32)
Sgs − Sgr Sgs − Sgr

where krl increases linearly from 0 to 1 in the range Slr ≤ Sl ≤ Sls, and krg increases linearly
from 0 to 1 in the range 1 - Sls ≤ 1 - Sl ≤ 1 - Slr.
Another set of linear functions that TOUGH2 can manage is:

k rl = k rl 0Sl (33)

k rg = 1 (34)

where krl0 has a constant value.

Corey’s curves
Corey’s curves can be written as follows (Corey, 1986):

k rl = (Sel ) 4 (35)

2
(
k rg = (1 − Sel ) 1 − (Sel )
2
) (36)

Grant’s curves
Grant (1977) reports krl as in Corey's model, whereas krg is calculated according to:

krg = 1 − krl (37)

Functions of Fatt and Klikoff


These expressions are reported in Fatt et al. (1959):

k rl = (Sel ) 3 (38)

k rg = (1 − Sel )
3
(39)

where Sel is calculated by Eq. (22) with Sls = 1.

van Genuchten – Mualem model


van Genuchten (1980) reported the following relationship between krl and h:

k rl (h ) =
{1− (αh) [1+ (αh) ] }
n −1 n m

[1 + (αh) ]
(40)
n m/2
A Modeling Approach Supporting Air Sparging System Design 95

where α, n, and m have been already used in the van Genuchten capillary function. Taking
into account Eq. (23), Eq. (25) and that θel = Sel:

2
⎧⎪ ⎡ ⎤ ⎫⎪
m

⎨1 − ⎢1 − (Sel ) m ⎥ ⎬
1
k rl (Sel ) = S 0 .5
el (41)
⎪⎩ ⎣ ⎦ ⎪⎭

This model showed a very good agreement with experimental data for sands and silty
loams, whereas less accurate results were obtained in describing the behavior of clays (van
Genuchten, 1980) and coarse-textured soils at low water content (Khaleel et al., 1995).
In TOUGH2 the following system of equations can be used:

⎧Eq.( 41) for S l < S ls


k rl = ⎨ (42)
⎩1 for S l ≥ S ls
and
⎧Eq.(37) for S gr = 0
k rg = ⎨ (43)
⎩Eq.(36) for S gr > 0

Function of Verma et al.


The model proposed by Verma et al. (1985) can calculate the relative permeability for
liquid and gas phases, according to Eq. (38) for krl vs Sl, and the following expression for krg:

k rg = A + B Sel + C (Sel )
2
(44)

where A, B and C are empirical constants. The authors measured their values for steam-water
flow in an unconsolidated sand, resulting in 1.259, -1.7615, and 0.5089 respectively.
Different values (1.3, -9.00, and 5.00 respectively) are reported in Scheid et al. (2003) for a
silty sand.

Case History
This section presents the modeling approach applied at a polluted site where an AS system is
operating, to assess the airflow path in soil in relation to the value of the process parameters
(AS injection and SVE extraction flowrates). Further simulations were carried out to study the
effect towards the airflow path of AS injection cycles and different plant configurations
(surface sealing, number and location of AS or SVE wells and well screen position), in order
to upgrade the system.

Site and Plant Configuration

The site is an industrial area, located in the northwest of Milan (Italy). Site stratigraphy [Fig.
6(a)] results in: a silty sand layer, with gravel and cobbles (Layer I), extending from ground
96 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

surface (g.s.) to the bottom of the phreatic aquifer (30 m below g.s.), followed by a clayey
sandy silt layer (Layer II) 2 m thick. Layer I is characterized by the following properties: f =
0.25, k = 3.6 10-10 m2. Wind evaporation method was applied to obtain the soil water retention
curve, whose data were fitted to obtain the van Genuchten parameter values (m = 0.715, and
α = 0.265 m-1) and residual water saturation (Slr = 0.12).

Figure 6. Site geometry (not in scale): vertical section (a) and horizontal section (b).

Groundwater flows to southeast; water table fluctuates between 13 and 18 below g.s.
during one year period, with a mean level of -15 m g.s.. According to the national regulation
in force when the characterization and the remediation activities started, soil is polluted by
monoaromatic solvents and light hydrocarbons, on a squared area of about 170 m2, -12 m g.s.
to -23 m g.s.; groundwater plume is 120 m long and 50 m large.
An AS system is installed at the site over a not paved area, including an injection well
PAS (external diameter Øext = 0.15 m) located in the centre of the horizontal section of the
polluted soil [Fig. 6(b)] and screened -22 m g.s. to -25 m g.s. [Fig. 6(a)]. A piezometer PZAS
is placed 6.25 m downgradient PAS [Fig. 6(b)] and screened -16 m g.s. to -25 m g.s.. A SVE
system is also installed, including a venting well PSVE, placed 2.75 m upgradient PAS [Fig.
6(b)] and screened -5 m g.s. to -12 m g.s. [Fig. 6(a)]; three monitoring wells (PZSVE1, PZSVE2,
PZSVE3) are located around PSVE at different distances (4.0 m, 6.0 m and 7.4 m respectively)
and directions [Fig. 6(b)], and screened as PSVE.
An AS pilot test was performed (no extraction at PSVE) using a mass flow rate QV,AS,pil =
5.33 10-3 kg s-1 (15 Nm3 h-1, referred at 0 °C and 1.013 105 Pa). Water mounding with respect
to natural conditions (Δhl) was recorded in PAS and PZAS after 60 minutes from the beginning
of the test and plotted versus the distance from PAS to calculate the amplitude of the Zone Of
Influence ZOIAS,pil = 8.0 m as linear extrapolation to Δhl = 0.
A SVE pilot test was also performed (no air injection at PAS) using a mass extraction rate
QV,SVE,pil = -4.27 10-2 kg s-1 (-120 Nm3 h-1). The air pressure drop ΔPg between natural and
A Modeling Approach Supporting Air Sparging System Design 97

PSVE operating conditions measured in PZSVE1, PZSVE2 and PZSVE3 was plotted as a function of
PZSVE distance from PSVE; ZOISVE,pil (about 18 m) was obtained as linear extrapolation of the
distance from PSVE where ΔPg = 0 Pa.
The full-scale treatment is being performed operating the systems at the pilot test flow
rates.

Modeling

The domain (horizontal section: 50 m x 50 m; height: 30 m) was discretized through an


irregular grid, whose X and Y directions are shown in [Fig. 6(b)]; cell dimensions ranged 0.10
m to 1.50 m along the vertical Z-axis, and 0.15 m to 9.2 m along the X- and Y- axes.
Two “extracells” outside the domain were linked to its bottom and top, in order to
simulate the aquitard and atmosphere, respectively. Each one has a volume of 1.0 1050 m3 and
the horizontal section as the domain.
Soil was modeled as an isothermal (T = 20 °C), homogeneous, isotropic medium, whose
total porosity, and intrinsic permeability values were assigned according to the
characterization performed at the site for Layer I. Eq. (27) and Eqs. (42), (43) were selected
for the capillary pressure and the relative permeabilities, respectively, whose parameter
values were those obtained for Layer I; small variation in these values is expected due to
variation in temperature (Narasimhan et al., 1977). For soil dry bulk density, a typical value
from the literature (1700 kg m-3) was assumed.
The “extracells” had the same physical properties of the domain, except that capillary
pressure was excluded. Air pressure Pg and air saturation Sg in soil under natural conditions
were obtained as stationary conditions of a simulation where the top “extracell” was fixed at
the atmospheric feature (Pg = 2.38409 105 Pa, Sg = 0.01) and the bottom at the hydrostatic
pressure (Pg = 1.01300 105 Pa, Sg = 0.99); the resulting groundwater level (-14.94 m g.s.) was
close to the average value measured at the site, with Sg ≤ 0.01 and Sl ≥ 0.99 beneath this level.
Source/sink wells of air at flow rate Q, were simulated as a finite number (N) of
positive/negative source cells of air, whose flow rate was set at Q/N; each cell had a squared
plane section (0.15 m x 0.15 m), and was 0.35 m long for the injection wells, and 1.0 m long
for the extraction wells. PAS was modeled as two cells located in the upper part of the screen,
due to air buoyancy limiting the injection depth; seven cells were located along the entire
length of the screen for PSVE. The numerical code calculates Ξg in Eq. (12) taking into
account the air mass flow and the volume of the cell. All unscreened parts of the wells were
simulated as disabled elements of the grid.
Surface sealing was modeled as a heterogeneity 0.10 m thick, and 20.3 m x 20.3 m wide
at the top of the domain, centered on the contaminated soil. The following properties were
assigned to this material: f = 0.01 (APAT, 2006), k = 1.0 10-100 m2, no capillary pressure, all
phases perfect mobile.
For simulations with source/sink cells, i.c. were loaded as natural conditions in soil; b.c.
at the top and at the bottom of the domain were set as in the simulation to generate natural
conditions in soil; the code assumes “no flow" b.c. for vertical boundaries and disabled cells.
98 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Model Verification

The model was verified by comparing pilot test simulations and field measurements.
Concerning AS test, field water mounding Δhl in PAS and PZAS was interpreted as water
pressure increase with respect to natural conditions at a depth equal to half of the screen wells
(-23.5 m g.s. for PAS, -20.5 m g.s. for PZAS); since no porous materials are present in wells, Pc
is negligible and Pl = Pg, therefore Δhl can be calculated from simulation results as the
variation between Pg at time of concern and natural condition, divided by (ρl g). Δhl resulting
from the model was plotted versus the distance from PAS; through linear extrapolation to Δhl
= 0, ZOIAS,pil from model resulted in 8.3 m [Fig. 7(a)], in agreement with field data.

Figure 7. Linear extrapolation of Δhl during the AS pilot test (a) and ΔPg during the SVE pilot test (b)
vs distance from injection/extraction well (>: value from the model at the monitoring wells; ---: linear
interpolation).

Concerning SVE pilot test, ΔPg was obtained as the difference between the Pg value at
stationary conditions under operation and Pg at natural conditions resulting at half depth of
the screen of the monitoring wells PZSVE (-8.5 m g.s.); these values were plotted as a function
of the distance of PZSVE from PSVE, and linearly extrapolated to ΔPg = 0 Pa, resulting in a
simulated ZOISVE,pil = 17.6 m [Fig. 7(b)], in agreement with field data.

AS and SVE Flow Rates

Fig. 8 shows air saturation at stationary conditions over a semi-section parallel to the XZ
plane passing through PAS, for simulations where PAS was active at different injection flow
rates (QV,AS: 5 to 20 Nm3 h-1). The isoline Sg = 0.01 (or Sl = 0.99) delimits water saturated
zone; Sg = 0.20 is the maximum value of gas saturation resulting from the simulations. The
dashed black line shows water table at natural conditions (no air injection), the brown one
delimits pollution in soil. As expected (USACE, 2008), as QV,AS grows the extension of the
zone nearby PAS partially saturated by air becomes larger; nevertheless, Sg values higher than
0.1 (black continue line) can be obtained only for QV,AS ≥ 10 Nm3 h-1 next to the screened part
of PAS.
A Modeling Approach Supporting Air Sparging System Design 99

Figure 8. Air saturation Sg at stationary conditions over a semi-section parallel to the XZ plane passing
through PAS [red thick line in (a)], for simulations where PAS is active at different injection volumetric
flow rates QV,AS [(b): 5 Nm3 h-1; (c): 10 Nm3 h-1; (d): 15 Nm3 h-1; (e): 20 Nm3 h-1]. The dashed black line
shows water table at natural conditions (no air injection), the brown one delimits pollution in soil, and
the black continue line the zone where Sg ≥ 0.1.

Figure 9. Air pressure Pg at stationary conditions over a semi-section that parallel to the XZ plane
passing through PSVE [red thick line in (a)], at different volumetric extraction flow rates QV,SVE [(b): -60
Nm3 h-1; (c): -90 Nm3 h-1; (d): -120 Nm3 h-1; (e): -150 Nm3 h-1). The dashed brown line delimits pollution
in soil.
100 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Fig. 9 shows air pressure Pg at stationary conditions over a semi-section parallel to the
XZ plane passing through PSVE, for simulations where PSVE was active at different extraction
flow rates (QV,SVE: -60 to -150 Nm3 h-1); the maximum value in the color scale is the air
pressure at water table. At the boundary on the right of each figure, Pg results in values as for
natural conditions (no influence of PSVE). As ⎪QV,SVE⎪ grows the depression cone next to the
screened part of PSVE enlarges, resulting in the minimum value (1.011 105 Pa) for the highest
absolute value flow rate which has been simulated [Fig. 9(e)]. For⎪QV,SVE⎪ ≥ 90 Nm3 h-1, a
depression zone is shown next to PSVE at ground surface (top boundary in the figures),
suggesting that atmospheric air is sucked into soil.
Fig. 10 shows air velocity vectors vg at stationary conditions when PAS and PSVE are
contemporarily operating as in the full-scale treatment performed at the site. SVE well (thick
green line on the left) sucks air from g.s. into the top of the screened part of the well and from
soil into the bottom part. The air injected by the AS well (thick orange line on the right)
flowing through the ZOTAS (Sg ≥ 0.1, within the white contour) is completely captured by the
extraction well, even if a small amount of air flowing through the ZOIAS (Sg ≥ 0.01, beneath
water table under natural conditions) escapes the collecting system.

Figure 10. Air velocity vectors (vg ≥ 1 10-4 m s-1) at stationary conditions when PAS and PSVE are
contemporarily operating (injection flow rate in PAS: QV,AS,pil; extraction flow rate in PSVE: QV,SVE,pil) over
an axonometry showing planes passing through the centre of the contaminated zone in soil [α and β in
(a)]. Planes γ and δ in (a) are parallel to the XZ plane and delimit pollution in soil along X direction. On
plane β, polluted soil is enclosed within the dashed brown line. ZOTAS (Sg ≥ 0.1) is enclosed within the
white oval.

Under the same operating conditions of Fig. 10, Fig. 11 shows results over the section
parallel to the Y’Z plane passing through PAS (thick orange line on the left) and PSVE (thick
green line on the right) [Fig. 11(a)]. In the lower part of Fig. 11(b), beneath water table under
natural conditions (triangle on the right boundary) Sg varies between 0.01 and 0.2. The white
oval shows the ZOTAS of PAS. In the upper part of Fig. 11(b), the air pressure Pg is shown as a
scalar ranging between 1.0115 105 Pa and the value at water table under natural conditions
A Modeling Approach Supporting Air Sparging System Design 101

(1.0150 105 Pa); one color step in the scale results in an air pressure variation of 25 Pa. Where
no perturbation occurs, isobars are horizontal lines; the zone enclosed within the inner closed
black line is the ZOTSVE for PSVE, where ΔPg ≥ 25 Pa, whereas pressure variations ΔPg with
respect to natural conditions are below 25 Pa in the zone enclosed between the outer closed
black line and the ZOTSVE. Water mounding, caused by air injection in PAS, is symmetrically
distributed around the injection well; however, asymmetry results in the ZOTSVE of PSVE. The
ZOTSVE of PSVE is smaller than the volume where air velocity vg is higher than 1 10-4 m s-1, but
overlaps the zone involved in water mounding. The vertical projection of the ZOTAS of PAS is
entirely enclosed in the ZOTSVE of PSVE based on ΔPg; nearly the total amount of air flowing
through the polluted area (dashed brown rectangle) falls within the zone around PSVE were vg
≥ 1 10-4 m s-1.

Figure 11. Air saturation Sg in the zone beneath water table under natural conditions (black triangle on
the right) and air pressure Pg in the zone above this level (b), over plane α parallel to the Y’Z plane (a)
passing through PAS and PSVE (operating at QV,AS,pil and QV;SVE,pil respectively) - steady state conditions.
Pg step in the color scale: 25 Pa. The white oval is the ZOTAS (Sg ≥ 0.1); ZOTSVE (ΔPg ≥ 25 Pa) is
enclosed within the inner closed black line. Air velocity vectors (vg ≥ 1 10-4 m s-1) are also shown. The
dashed brown rectangle delimits pollution in soil.

AS Injection Cycles

In order to evaluate the opportunity of using a pulsed air sparging to enhance groundwater
mixing, the air plume evolution during the transient state was assessed. Fig. 12 and Fig. 13
show the air pressure Pg and air saturation Sg over the semi-section of Fig. 8(a) as a function
of time when PAS is operating at the injection flow rate QV,AS,pil. Fig. 12 suggests water
mounding during the transient state, whose duration is about 120 min; after this time [Fig.
12(e)], steady state conditions establish. The ZOIAS around PAS during the transient state (Sg ≥
0.01 in Fig. 13) is smaller than the zone involved in water mounding (Fig. 12), because the
injected air spreads laterally when it reaches the capillary fringe. Fig 13 also shows that the
injected air creates an unsaturated reversed cone shaped zone around the injection well,
whose size gets wider in time. Approximately 2 hours after the air injection has been turned
on, the ZOIAS (Sg ≥ 0.01) and the ZOTAS (Sg ≥ 0.1) do not change significantly over time, as
102 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

results by comparing Fig. 13(e) to Fig. 8(d). The optimal duration of the “on” period for an
AS cycle is about 2 hours, since a shorter time does not result in the maximum ZOIAS,
whereas a longer time does not promote water mixing. Shutting off the air injection results in
a significant decrease (Sg ≤ 0.05) in the degree of saturation to air in the cone within 10
minutes (pictures not shown), suggesting that a pulsed air sparging is not suggested, as
concerns the air flow problem.

Figure 12. Air pressure Pg over a semi-section parallel to the XZ plane passing through PAS [as in
Figure 8(a)] operating at QV,AS,pil, at different times from the beginning of the air injection 2 min (a), 8
min (b), 16 min (c), 30 min (d), and 120 min (e). Pg ranges from the value at the bottom of the domain
and the value at water table under natural conditions (black triangle on the left of each picture).

Figure 13. Air saturation Sg over the semi-section of Figure 8(a) at times: 2 min (a), 8 min (b), 16 min
(c), 30 min (d), and 120 min (e).

Plant Configuration

Screen location

Simulations were carried out to study the effect towards the airflow path in soil due to the
screen position. Screen position was not varied contemporarily for PAS and PSVE, as airflow
behavior under water table does not significantly affect air flow in the vadose zone.
For the AS well, a configuration was simulated where the top of the screen was located at
the bottom of contamination in soil (source cells positioned -23 m to -23.7 m g.s.). For the
simulated flow rates (as in Fig. 8), the ZOIAS extends in the contaminated soil just a little bit
more than for the configuration installed at the site, but the overlapping between pollution in
soil and the ZOTAS decreases because the ZOTAS also extends beneath the polluted soil
(picture not shown).
A Modeling Approach Supporting Air Sparging System Design 103

Also for SVE well, the screen position was shifted downwards (-6 m g.s. to -13 m g.s.)
with respect to the current position in order to partially locate it (for about 1 m) within the
contaminated soil. For the simulated flow rates (as in Fig. 9), a reduced sucking of
atmospheric air into soil, and a decreased overlapping between the ZOTSVE and the polluted
soil are obtained. Fig. 14 shows, just as examples, the results at stationary conditions on semi-
sections parallel to the XZ plane passing through the PAS [Fig. 14(a)], to be compared to Fig.
8(d) or PSVE [Fig. 14(b)], to be compared to Fig. 9(d); the dashed brown line delimits
pollution in soil.

Figure 14. Sg (a) over a semi-section parallel to the XZ plane passing through PAS [as in Figure 8(a)] and
Pg (b) over a semi-section parallel to the XZ plane passing through PSVE [as in Figure9(a)], operating at
QV,AS,pil and QV,SVE,pil respectively - steady state conditions. PAS is screened -23 to -26 m g.s. and PSVE -6
to -13 m g.s.. The dashed black line in (a) shows water table under natural conditions; the brown dashed
line delimits pollution in soil.

Figure 15. Air pressure Pg at steady state over a section parallel to the XZ plane passing through PSVE
[red line in (a)], operating at QV,SVE,pil. The dashed green zone in (a) and the horizontal green line in (b)
show the surface sealing. The brown dashed line in (a) delimits pollution in soil. Each color step results
in 25 Pa. Air velocity vectors with vg ≥ 1 10-4 m s-1 are also shown.
104 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Surface sealing

Fig. 15(b) shows results at stationary conditions about the simulation where the surface
sealing was applied, and only PSVE was operating with an extraction flow rate QV,SVE,pil. Air
pressure range in the color scale is set to let a direct comparison with Fig. 9(d), even if lower
values (up to 1.0113 105 Pa) are obtained with sealing. Atmospheric air is sucked no more
into soil; beneath sealing (continuous horizontal green line), air flows along X direction,
without crossing the impermeable layer. At half depth of the screen (-8.5 m g.s.) in the
monitoring wells PZSVE1, PZSVE2 and PZSVE3 [Fig. 16(a)], the air pressure [Fig. 16(c)] is lower
than without sealing [Fig. 16(b)]; at this depth, the amplitude of ZOTSVE (ΔPg ≥ 25 Pa) is 5.1
m, about twice the value obtained without sealing (2.6 m).

Figure 16. Air pressure Pg at steady state over a section parallel to the XY plane passing at half depth of
the screen in the monitoring wells PZSVE1, PZSVE2 and PZSVE3 [(a)], when PSVE is operating at QV,SVE,pil.:
(b) without surface sealing; (c) with surface sealing.

Figure 17. Air saturation Sg at stationary conditions over horizontal sections parallel to the XY plane
(a). The simulated plant is composed by four AS wells, each one operating at QV,AS,pil, screened as PAS,
and 4.5 m far. Sections show results at different planes: α at -23 m g.s. (b), β at -22.3 m g.s. (c), γ at -
18.5 m g.s. (d) and δ at -15.0 m g.s. (e). The dashed brown rectangle delimits pollution in soil.
A Modeling Approach Supporting Air Sparging System Design 105

Number and location of AS or SVE wells

Simulations were performed to assess the proper number and location of AS and SVE
wells in order to entirely cover the polluted zone in soil by the injected air. Plants with two,
three or four injection wells were modeled, each one screened -22 m to -22.7 m g.s. and
operating at QV,AS,pil. The optimal solution, composed by four injection wells placed 4.5 m
from the others, is reported in Fig. 17, where the ZOIAS at stationary conditions over
horizontal sections parallel to the XY plane at different depths [e.g.: at the bottom of pollution
in soil, that is -23 m g.s., in Fig. 17(b); at the middle of the active zone of the screen of
injection wells, that is -22.35 m g.s., in Fig. 17(c); at half depth between water table under
natural conditions and the top of the screen of the injection wells, that is -18.5 m g.s., in Fig.
17(d); at water table under natural conditions, that is -15 m g.s., in Fig. 17(e)] are shown .
Asymmetry in the pictures is due to the grid, whose number of elements is limited by
PetraSym.
Fig. 18 compares air saturation obtained for plant configurations where four wells are
place at a distance of 2 m [Fig. 18(a)] and 4.5 m [Fig. 18(b)]. In the upper part of the figure,
sections parallel to the XZ plane passing through the baricentric position of the AS injection
system are shown, whereas in the lower part, sections parallel to the XY plane at -18.5 m g.s.
are shown [plane γ in Fig. 17(a)]. Where a reduced distance among wells is applied, values of
Sg ≥ 0.1 can be obtained within the red circle but air saturation is low over a large part of the
polluted soil. With a longer distance among wells, Sg ≥ 0.05 within the orange circle, wider
than in Fig. 18(a).

Figure 18. Air saturation Sg at stationary conditions for an injection system composed by four wells, 2
m far (a) or 4.5 m far (b). In the upper part, plane α is parallel to the XZ plane (passing through the
baricentric position of the AS injection system). In the lower part, plane β is parallel to the XY plane at
-18.5 m g.s.; red and orange circles enclose zones where Sg ≥ 0.1 and Sg ≥ 0.05 respectively. The dashed
brown line delimits pollution in soil.
106 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

Figure 19. Plane section of a possible upgraded system, composed by 4 AS wells (PAS1, PAS2, PAS3, and
PAS4) and 4 SVE wells (PSVE1, PSVE2, PSVE3, and PSVE4). The dashed brown line delimits pollution in soil.

Figure 20. Results at stationary conditions, over planes shown in (a) or (b), concerning a possible upgraded
system (as in Figure 19), composed by 4 injection wells PAS (each operating at QV,AS,pil) and 4 extraction wells
PSVE (each operating at QV,SVE,pil): (c) axonometric view of air saturation Sg; (d) axonometric view of air
pressure Pg; (e) Pg over plane η, passing at half depth of the screen of the SVE extraction wells. Dashed
brown lines delimit pollution in soil. Air velocity vectors (vg ≥ 1 10-4 m s-1) are also shown.
A Modeling Approach Supporting Air Sparging System Design 107

Figure 21. Air pressure Pg at steady state for the plant configuration of Figure 19 with surface sealing as
in Fig 15(a). PAS (1 to 4) and PSVE (1 to 4) are operating at QV,AS,pil, and QV,SVE,pil respectively: (a)
axonometric view [as in Figure 20(d)]; (b) over plane η [as in Figure 20(e)]. The dashed brown line
delimits pollution in soil. Each color step results in 25 Pa; air velocity vectors (vg ≥ 1 10-4 m s-1) are also
shown.

Other simulations were carried out for plant configurations composed by 4 injection wells
at a distance of 4.5 m from each other, and 1 to 5 SVE wells, in the attempt to capture all the
air flowing through the polluted soil. The suggested system is composed by 4 AS wells (4.5
m far) and 4 SVE wells (10.5 m far) placed as shown in Fig. 19. Fig. 20(c) shows an
axonometric view with the plane parallel to XZ passing through PAS1 and PAS2 (α), the plane
parallel to YZ passing through PSVE2 and PSVE3 (β), and planes γ and δ delimiting pollution in
soil. Over plane α, air velocity vectors with vg ≥ 1 10-4 m s-1 are shown, resulting in the
capture in the vadose zone of air flowing through polluted soil under the water table. In the
axonometric view shown in Fig. 20(d), Pg ranges from the maximum value (1.0150 105 Pa) at
the water table under natural conditions, to the minimum value (1.0113 105 Pa) in the middle
of the screened part of SVE wells; plane ζ leans against the top of the polluted soil, where a
cross-shaped zone in the core of the polluted soil is not enclosed within the ZOTSVE (ΔPg ≥ 25
Pa) of the upgraded plant. Fig. 20(e) shows plane section η [Fig. 20(b)], where isolines
suggest that the ZOTSVE of the upgraded plant is wider than in Fig. 16(b) (one extraction
well); however, a certain amount of polluted area is still outside. Including a surface sealing
as in Fig. 15(a), the ZOTSVE extends to cover the entire polluted area [Figs. 21(a) and 21(b)].

Conclusion
Simulations about the AS plant as actually installed at the site suggest that one injection well
ensures air saturation values high enough for an effective treatment only over a very small
volume of the polluted soil. As a matter of fact, the duration of the treatment estimated during
the system design has been widely exceeded. A higher injection flow rate at PAS does not
result in a decisive increase of air saturation within the polluted soil, also due the low intrinsic
permeability value. The SVE system is able to collect air flowing through the zone with high
108 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

air saturation values around PAS, but a certain amount of air flowing through pollution in soil
escapes the collecting system.
A pulsed AS does not seem to be suitable for the site. A variation in the screen position
of the AS wells is not advisable, due to a reduced overlapping of the ZOTAS and pollution in
soil. Surface sealing significantly increases the amplitude of the ZOTSVE of the SVE system. A
possible system upgrade is suggested, composed by 4 AS wells and 4 SVE wells, each
operating as during the full-scale treatment performed up to now by PAS and PSVE, and surface
sealing.

List of Symbols
A empirical constant (-)
B empirical constant (-)
C empirical constant (-)
CS specific heat of the rock (J kg-1 K-1)
ez unit vector in z direction
F mass (kg m-2 s-1) or heat flux (J m-2 s-1)
f soil porosity (-)
fs safety factor (-)
g gravitational acceleration (m s-2)
H Henry’s law constant (Pa m-3 mol–1)
h pressure head (m)
ha air-entry pressure head or bubbling pressure head (m)
k intrinsic permeability of soil (m2)
kr relative hydraulic conductivity (-)
krl0 constant (-)
m empirical coefficient (-)
M mass per volume (kg m-3) or energy per volume (J m-3)
n empirical coefficient (-)
n normal vector on surface element Γj
P pressure (Pa)
Pc capillary pressure (Pa)
Pcr minimum value of Pc (Pa)
Pf overburden pressure (Pa)
Ph hydrostatic pressure (Pa)
Pi injection pressure (Pa)
Pmax maximum injection pressure (Pa)
Pmin minimum injection pressure (Pa)
q Darcy velocity (m s-1)
QV airflow rate (m3 s-1)
r source or sink (kg m-3)
R universal gas constant (Pa m3 mol-1 K-1)
S saturation (-)
Sel effective liquid saturation (-)
Sr residual saturation (-)
A Modeling Approach Supporting Air Sparging System Design 109

Ss maximum saturation (-)


T temperature (K)
t time (s)
u specific internal energy (J kg-1)
V volume (m3)
vg air velocity vector (m s-1)
X mass fraction (-)
ZOI Zone Of Influence
ZOT Zone Of Treatment
ztws depth of the well screen top from ground surface (m)
zwc difference between depth to the top of PAS well screen and pre-sparging depth to
the free-water surface within the sparging well (m)

Greek letters

Δh l water mounting (m)


ΔP pressure difference (Pa)
Γ closed surface boundary of an arbitrary subdomain (m2)
Ω domain
Ξ sink or source (kg m-3 s-1)
Ψ suction (m)
α empirical coefficient (m-1)
χ fitting parameter (-)
ε fitting parameter (-)
μ dynamic viscosity (Pa s)
θ volumetric phase content (-)
θel effective volumetric liquid content (-)
θr residual volumetric phase content (-)
θs maximum volumetric phase content (-)
ρ density (kg m-3)
σ surface tension of water
ωg air molecular weight (g mol-1)

Subscripts

0 to identify a specific value or a fitting parameter


AS air sparging
c capillary
e effective
f pore
g air
l water
n integer
110 Sabrina Saponaro, Sara Puricelli, Elena Sezenna et al.

r residual
SVE soil vapor extraction
β phase
κ mass or heat component

References
APAT Criteri metodologici per l’applicazione dell’analisi assoluta di rischio ai siti
contaminati, Rev.1; Agenzia per la Protezione dell’Ambiente e per i Servizi Tecnici,
2006; p 47.
ASTM Standard guide for measuring matric potential in vadose zone using tensiometers; D
3404-91, 2004.
ASTM Standard test methods for determination of the soil water chararcteristic curve for
desorption using a hanging column, pressure extractor, hilled mirror hygrometer, and/or
centrifuge; D 6936-02; 2003.
Berkowitz, B. ; Silliman, S. E.; Dunn, A. M. Vad. Zone J. 2004, 3, 534-548.
Byrnes, A. P. (2008). Issues with gas and water relative permeability in low-permeability
sandstones
[http://www.searchanddiscovery.net/documents/abstracts/2005hedberg_vail/abstracts/ext
ended/byrnes/byrnes.htm].
Corey, A. T. Mechanisms of immiscible fluids in porous media; 86-50589; Water resources
publications: Littleton. CO, 1986.
Cornelis, W. M.; Khlosi, M.; Hartmann, R.; van Meirvenne, M., De Vos, B. Soil Sci. Soc. Am.
J. 2005, 69, 1902-1911.
Dixon, K. L.; Nichols R. L. Soil vapor extraction system design: a case study comparing
vacuum and pore gas velocity cut off criteria; in conjunction with Contract No. DE-
AC09-96SR18500; U.S. Department of Energy: Aiken, SC, 2006.
EOLBNL Tough2 User’s Guide - Version 2.0; LBNL-43134: Earth Sciences Division,
Lawrence Berkeley National Laboratory: Berkeley, CA, 1999.
Fatt, I.; Klikoff, W. A. AIME Trans. 1959, 216, 426-432.
Garnier, P.; Ezzine, N. ; De Gryze, S.; Richard, G. Vad. Zone J. 2004, 3, 716-721.
Grant, M. A. Proceeding of ASME/AIChE Heat Transfer Conference; paper 77-HT-52: Salt
Lake City, UT, 1977; 15-17.
ISO Soil quality - Determination of pore water pressure - Tensiometer method; 11276:1995
(E), 1995.
ISO Soil quality - Determination of the water-retention characteristic - Laboratory methods;
11274:1998 (E), 1998.
ISO Soil quality - Determination of unsaturated hydraulic conductivity and water-retention
characteristic - Wind's evaporation method; 11275:2004 (E), 2004.
ISO Soil quality - Sampling - Part 4: Guidance on the procedure for investigation of natural,
near-natural and cultivated sites; 10381-4:2003 (E), 2003.
Khaleel, R.; Relyea, J. F.; Conca, J. L. Water Resour. Res. 1995, 31, 2659-2668.
Luckner, L.; van Genuchten, M. Th.; Nielsen, D.R. Water Resour. Res. 1989, 25, 2187-2193.
A Modeling Approach Supporting Air Sparging System Design 111

Lundegard, P. D. In situ aeration: air Sparging, bioventing, and related remediation


processes. Hinchee, R. E.; Miller R. N.; Johnson P. C. Ed.; Battelle Press: Columbus,
OH, 1995; pp 21-30.
Lundegard, P. D.; LaBrecque D. J. Cont. Hydro. 1995; 19, 1-27.
McCray, J. E.; Falta R.W. J. Contam. Hydrol. 1996, 24, 25-52
Meadows, D. G.; Young, M. H.; McDonald, E. V. Soil Sci. Soc. Am. J. 2005, 69, 807 – 815.
Mei, C. C.; Cheng, Z.; Ng, C. O. Appl. Math. Modelling 2002, 26, 727-750
Milly, P. C. Water Resour. Res. 1982, 18, 489-498.
Moseley, W. A.; Dhir, V. K. J. Hydrolo. 1996, 178, 33-53.
NFESC Final air sparging guidance document; Technical Report TR-2193-ENV; Battelle for
Naval Facilities Engineering Service Center: Port Hueneme, CA, 2001.
Pickens, J. F.; Gillham, R. W.; Cameron, D. R. J. Hydrolo.1979, 40, 243-264.
Scheid, Y.; Semprich, S.; Chinkulkijniwar, A. Proceeding of TOUGH Symposium: Berkeley,
CA, 2003; 1-6.
Slawinsky, C.; Witkowska-Walczak, B.; Walczak, R. T. In Determination of water
conductivity coefficient of soil porous media; Slawinsky, C.; Walczak, R. T. Ed.; Institute
of Agrophysics PAS: Lublin, PO, 2004.
Suthersan, S. S. Remediation engineering: design concepts; CRC Press LLC: Boca Raton,
FL, 1997; pp 91-122.
Thomson, N. R.; Johnson, R. L. J. Hazard Mater. 2000, 72, 265-282.
Udell, K. S.; Fitch, J. S. Proceeding of 23rd ASME/AIChE National Heat Transfer
Conference: Denver, CO, 1985.
University of Wisconsin UTTU – Underground tank technology update 2000, 14, 2-5.
USACE Engineering and design – soil vapor extraction and bioventing; EM 1110-1-4001;
Department of the army, US Army Corps of Engineers: Washington, DC, 2002.
USACE In-situ air sparging engineer manual; EM 1110-1-4005; U.S. Army Corps of
Engineers: Washington, DC, 2008.
USEPA Development of recommendations and methods to support assessment of soil venting
performance and closure; EPA/600/R-01/070; U.S. Environmental Protection Agency:
Washington, DC, 2001.
USEPA How to evaluate alternative cleanup technologies for underground storage tank sites
– a guide for corrective action plan reviewers; EPA/510-R-04-002; U.S. Environmental
Protection Agency: Washington, DC, 2004; pp VII 1-28.
USGS (2008). Soil water retention measurement [http://wwwrcamnl.wr.usgs.gov/uzf/
retention.html].
Van Dijke, M. I. J.; van der Zee S. E. A. T. M.; van Duijn, C. J. Adv. Water Resour. 1995, 18,
319-333
van Genuchten, M. Th. Soil Sci. Soc. Am. J 1980, 44, 892-898.
van Genuchten, M. Th.; Nielsen, D.R. Ann. Geo-phys. 1985, 3, 615–628.
Verma, A. K.; Pruess, K.; Tsang, C. F.; Witherspoon, P. A. Proceeding of 23rd ASME/ AIChE
National Heat Transfer Conference: Denver, CO, 1985; 135-143. a)
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 113-131 © 2008 Nova Science Publishers, Inc.

Chapter 3

GROUNDWATER INTERACTIONS WITH SURFACE


WATERS: CONSEQUENCES ON DIFFUSE POLLUTION
PATHWAYS

A. Azzellino, R. Salvetti and R. Vismara


D.I.I.A.R. - Environmental Section - Politecnico di Milano,
Piazza Leonardo da Vinci, 32 - 20133 Milano, Italy

Abstract
The interactions between groundwater and surface water are complex. Surface-waters
and groundwaters are, in fact, linked components of a hydrologic continuum. In general,
diffuse pollution in surface waters is difficult to quantify since it follows a multitude of
pathways and acts on different time scales. During rainfall events most of the diffuse pollutant
load follows the surface runoff pathways and, reaches the surface acquifers however, a
fraction of this load will follow the sub-surface runoff pathways and it will possibly reach the
surface acquifers after a certain time lag. The time scale of the sub-surface runoff pathways is
very different from the surface runoff time scale and rarely a subsurface diffuse pollution
event can be directly correlated to a specific rainfall event. This is the reason why even though
there are models that enable to simulate the groundwater-surface water system (GW-SW), yet
the effect of these interactions in terms of diffuse pollution pathways and their correspondent
effect on the quality of surface waters to date are largely unknown. To upgrade the conceptual
modeling of the “groundwater–surface water” system, a broader perspective of such
interactions across and between surface water bodies is needed. Multidimensional analyses
may help in understanding the effect of such interactions, as the characterization of the
hydraulic interface and its spatial variability.
To fully understand these interactions, modeling studies need to be coupled to sound and
robust monitoring of surface- and ground- water quality data. Modeling can be combined with
multivariate statistical techniques (e.g. factor analysis) to improve our capability to “detect”
the effect of the sub-surface runoff on the water quality of specific water courses. With studies
of this kind, it was proved that the sub-surface runoff can be a significant source of diffuse
pollutant even in dry weather conditions. Studying nitrate concentrations in different lowland
rivers, we found that up to 60% of the measured load could be apportioned as sub-surface-
runoff-derived.
Notwithstanding, to date, very few studies have attempted to overlay measurements with
the conceptual modeling of the GW-SW system. In this commentary, studies about the GW-
114 A. Azzellino, R. Salvetti and R. Vismara

SW system will be reviewed, the key aspects and the research needs and challenges facing this
evolving field will be discussed outlining how site-to-region regionalization studies and cross-
disciplinary collaborations could help in improving the understanding of the complexity of the
GW-SW system.

Introduction
Surface-water bodies are integral parts of groundwater flow systems. Groundwater interacts
with surface water ranging from small streams, lakes, and wetlands in headwater areas to
major river valleys and seacoasts. Although it is generally assumed that topographically high
areas are groundwater recharge areas and topographically low areas are groundwater
discharge areas, this is true primarily for regional flow systems. The interactions of streams,
lakes, and wetlands with groundwater are, in fact, affected by the positions of the water
bodies with respect to groundwater flow systems, geologic characteristics of their beds, and
their climatic settings. If the movement of surface water and groundwater is controlled to a
large extent by topography and the geologic framework of an area (i.e. physiography), the
sources of water to, and losses of water from, the earth’s surface are controlled by climate
(Winter, 1999). Therefore, it is necessary to understand the effects of physiography and
climate on groundwater flow systems in order to fully understand the interaction of
groundwater and surface water.
Because of the interchange of water between these two components of the hydrologic
system, development or contamination of one commonly affects the other (Winter et al. 1998;
Rozemeijer and Broers, 2007).
Hydrologic processes associated with the surface water bodies themselves, such as
seasonally high surface-water levels and evaporation and transpiration of groundwater from
around the perimeter of surface water bodies, are a major cause of the complex and seasonally
dynamic groundwater flow fields associated with surface water.
In such a complex system the thorough understanding of each component is needed for
effective management of water resources. Groundwater flow systems may be defined by the
boundary conditions imposed by their physiographic framework and by the distribution of
recharge. In the simplest framework, a rectangular aquifer is bounded by no-flow boundaries
at its base and on one side, surface water fully penetrates the aquifer on the other side, and
internally it is isotropic and homogeneous. Flow from the aquifer to the stream in such a
setting is largely one dimensional following a pulse of recharge uniformly distributed across
the aquifer’s upper boundary. However, natural groundwater systems generally do not have
these simple boundary conditions and are not composed of isotropic and homogeneous
porous media. To address the need to understand more realistic hydrologic systems, several
studies have been conducted that evaluated the effects of geologic framework on regional
groundwater flow systems (de Lange, 1999; Hudak, 2000; Lambs L. 2004; Goni, 2006).
These studies gave insights not only for the understanding of regional groundwater flow
systems, but also of how such flow systems interact with surface water.
Some studies showed that the chemistry of lake waters reflects the magnitude of the
groundwater flow systems that discharge to them (e.g. Wollschläger et al., 2007).
Additionally, long-term studies of wetlands present in lake area indicated that the major-ion
water type can change with major variations in climate: wetlands that normally contain
bicarbonate water can become dominated by sulfate water during major droughts, and
Groundwater Interactions with Surface Waters 115

wetlands having sulfate water can become dominated by bicarbonate water during major wet
periods (LaBaugh et al. 1996). Winter and Rosenberry (1998) documented substantial
changes in the relation of these wetlands to groundwater that resulted from century-scale
periods of drought and precipitation.
Streams are not only the terminal points of groundwater flow and the start of the surface
water system, but they are also critical components of the riparian and riverine ecology
(Woessner, 2000). The development and the modification of riparian woodland often reflect
the evolution of hydrological events (Tabacchi et al., 2000). The fluvial plain is, in fact, a
point of complex interaction between streams and the groundwater system (Lambs, 2000).
Stream and riparian ecologists have cited the importance of the mixing of stream water and
groundwater, and refer to the zone in which this occurs as the hyporheic zone. Mixing of
surface water and groundwater takes place within the upper layers of the channel sediments.
Such near-channel exchange occurs at many scales, from centimetres to tens of metres
depending on the bed geometry and the hydraulic-potential strengths. Pool and riffle
sequences characterize many high-gradient streams. It has often been found that surface water
enters channel sediments at the head of riffles and exits at the riffle base in pools. Water may
also circulate out of the stream as it enters the riffles, flow through the adjacent banks and
back into the down-gradient pools (Woessner, 2000). The physical characteristics of
groundwater and surface water have been considered to be distinctive, although more recent
studies have described surface water as a ‘perched groundwater aquifer’, in recognition of the
isotopic history of both components. Despite this, it is generally considered that groundwater
is characterized by stable flow, even temperature and a stable chemical composition that
reflects the underlying aquifer geology. In applied hydrology, the role of groundwater in
sustaining low flows has provided the focus for research (Sear et al., 1999). Stable isotopes,
such as 18O, are useful tools that allow water movement to be traced (Guay et al., 2004;
Lambs, 2004; Goni, 2005; Kattan, 2006). Two main water sources are typically present: (i)
river water, depleted of heavy isotopes, originating in the mountains, and (ii) groundwater,
which gives for temperate climate an annual average of the local rainfall. River water levels
and topographical details are certainly the main causes of the inflow of groundwater through
the river bank (Lambs, 2000; Lambs et al., 2002).

Groundwater and Diffuse Pollution


The heavy use of nitrogen (N) fertilizer for intensive farming and cropping systems with low
N use efficiency have been often responsible for nitrate overloading into groundwater (Strebel
et al., 1989). Since the 1960s, nitrate loading in both surface water and groundwater has
drawn great attention worldwide. Nitrate pollution cases that exceeded the threshold as
recommended by the World Health Organization (50 mg NO3- l-1) and by the USA (10 mg N
l-l) for drinking water were reported in many countries (Strebel et al., 1989; Fried, 1991;
Nolan et al., 1997; Hudak, 2000; Agrawal et al., 1999). Therefore the agriculture-derived
nitrate pollution of groundwater has become an environmental issue (Spalding and Exner,
1993) which may be the cause of algal blooms and eutrophication in aquifers, and it could
even produce potential hazards to human health (Knobeloch et al., 1992; Fan and Steinberg,
1996; Gelberg et al., 1999; Gulis et al., 2002) though such a link is not fully proven and still
under dispute (Forman et al., 1985; Van loon et al., 1998). Non-point source pollutants such
116 A. Azzellino, R. Salvetti and R. Vismara

as nutrients, pesticides, heavy metals, and sediments are transported from the land by surface
water and groundwater pathways and, typically, pollutant sources cannot be attributed to one
particular discharge location rather to a large area (Nikolaidis et al., 1998). For this reason,
Mutch (1998) suggested considering the sample location within a watershed and the scale of
water flow when trying to interpret the nitrate data and to assess the success of Best
Management Practices (BMPs). It was observed that in the regions where N fertilizer
application rates are above 500 kg N ha-1 and N use efficiency is less than 40% the nitrate
concentration in groundwater may usually exceed the standard for drinking water (Zhang et
al., 1996). Irrigation may further enhance nitrate leaching to aquifers from the farmlands
(Singh et al.,1995; Hudak, 2000). Other factors such as land use types, vadose zone thickness,
soil texture, timing of fertilizer application, and land use types may have too significant
impacts on N loading in shallow groundwater (Mclay et al., 2001). This might be the
consequence of the non-linear spatial variation of the hydraulic and geological heterogeneity
present within a watershed (Nikolaidis et al., 1998; Townsend et al., 1996).
Observed pollution may be also the consequence of farming practices many years ago
rather than current practices (Singh and Skelon, 1978; Mutch, 1998). The pollutant source
may not spatially and temporally coincide with the polluted site.
Liu et al. (2005), through a monitoring study carried out in 1998/1999, showed that
nitrate-N concentrations in groundwater may fluctuate seasonally. Nitrates may reach a peak
in some seasons and drop to the lowest level in other seasons. Moreover, they observed that
agriculture-derived nitrate-N pollution may develop non-uniformly even in a relatively
uniform farming area. They did not find, in fact, a clear correlation between N application
rates for different cropping systems and nitrate-N concentrations in the underlying
groundwater. Their finding suggests that nitrate-N levels in groundwater should be closely
related to the geographical location of the system instead of the specific system management
including N rates applied. The regional differentiation of non-point source pollution of
agriculture-derived nitrate-N in fact may play a major role in groundwater pollution.

Groundwater Pollution and Surface Waters


Pollution of groundwater and surface water by diffuse sources may be a serious problem
(Bellos et al., 2004; Campling et al., 2005). Use of fertilizer and animal manure in
agricultural areas is the most important non-point source in many rural catchments (Cinnirella
et al., 2005; Oenema and Roest, 1998; Schreiber et al., 2001; Wood et al., 2005). Discharge
of groundwater from agricultural land towards the surface water system may cause
exceedance of water quality standards (Van der Molen et al., 1998; Oenema and Roest, 1998;
Oenema et al., 2005) or simply produce additional loads (Azzellino et al., 2006).
Ground- and surface water management and monitoring in many countries is coordinated
by different authorities operating in different geographical sub-regions. Traditionally, in fact,
the monitoring and the management of groundwater and surface-water resources and quality
have been largely undertaken in isolation. Surface-water management programmes consider
the quality of urban rivers and direct discharges to them, e.g., industry pipe-end or sewage
discharges, but often fail to examine the relevance of quality within underlying groundwaters
or surrounding land. Clearly relationships exist and contaminated land within conurbations
may leach and cause pollution of underlying groundwaters that laterally flow and discharge as
Groundwater Interactions with Surface Waters 117

baseflows to surface waters (Shepherd, et al. 2006). Recent legislation is now requiring more
integrated environmental understanding, both in relation to risk assessment of contaminated
land (EA, 1999) and management of the water environment, in particular the European
Community Water Framework Directive (WFD) (Council of Europe, 2000). The WFD
requires that quality-flow compliance at a particular surface-water reach entail consideration
of all upstream inputs, including contaminated land and groundwater contributions. Besides
European Union, many other governments are now stimulating a more integrated approach to
managing the soil-groundwater-surface water system. In US, for instance, there is no
uniformity. Some states have some form of coordinated management to prevent harm to
surface water rights holders from groundwater pumping, while other states do not. The
management systems range from completely unified water management statewide, with all
forms of water treated as part of a common source, to limited management only in special
parts of the state (Peck, 2007). Besides the difficulties of changing the focus of the water
legislation, it must be stressed that diffuse pollution’s effect on surface and groundwaters is
often difficult to quantify. Diffuse pollutants, in fact, enter the environment through a
multitude of pathways at different temporal scales. During rainfall events most of diffuse
pollutant loads follow the surface runoff pathways and, at the net of the plant uptake, reach
the aquifers. On the other hand, a fraction of these loads follows the sub-surface runoff
pathways and may possibly reach the surface waters after a certain time lag. Very rarely the
sub-surface pollution events can be directly correlated to a specific rainfall event.

Source Apportionment Methods for the Surface Waters-


Groundwater System
Source apportionment is the estimation of the contribution from different sources to pollution.
Source apportionment deals with the pollution loads actually entering the aquatic
environment, as opposed to raw emissions (e.g. the agricultural nutrient loss from the root
zone or household wastewater entering the sewerage system, before purification). Generally,
a distinction is made between:

• point sources such as discharges from urban wastewaters, industries;


• diffuse sources including background losses (natural land, for example forest), losses
from agriculture, losses from scattered dwellings and atmospheric deposition on
water bodies.

Point sources are defined as stationary locations or fixed facilities from which pollutants
are discharged. The discharges are often monitored at the outlet from a wastewater treatment
plant, but may also be estimated based on information on the number of population
equivalents connected to a wastewater treatment plant and the type of wastewater treatment.

Diffuse losses are pollution from widespread activities with no specific point of
discharge, such as losses from natural areas and agricultural land, losses from paved areas,
etc.
118 A. Azzellino, R. Salvetti and R. Vismara

When estimating nutrient inputs to a river catchment or to the sea, two approaches can be
used (EEA Report, 2005):

9 a load-oriented approach, where the diffuse loss is estimated as the difference


between the total load measured at a river monitoring station and the measured
emissions from point sources upstream of the monitoring station. Estimates of
retention and losses in the river system are added to calculate the losses at source
(before retention).

Diffuse sources = CatchmentMeasured load river station - Point sourcesMeasured load (+ retention and
losses in river system);

9 A source-oriented approach, where the diffuse losses are estimated using export
coefficients from catchments with similar characteristics. The natural background
loss can be estimated using export coefficients from undisturbed catchments and the
agricultural loss can be estimated using export coefficients from catchments with
similar agricultural characteristics. Estimates of retention and losses in the river
system can be subtracted to calculate the total load at the river mouth (after
retention).

Total sources = Point sourcesMeasured/Estimated load + Diffuse sourcesEstimated load (– retention


and losses in river system)

In both approaches, the point sources are considered from a source-oriented point of
view, using measured discharges or sometimes standards for per capita discharges. The main
difference between the two approaches is the estimation of diffuse sources.
Within these two basic approaches, there are several ways to do the calculations.
Schoumans and Silgram (2003) made a review of different types of quantification tools for
nutrient losses to rivers. These quantification tools were established for different regions and
different tasks. They differ in their complexity and their resolution in time and space, and
they need different levels of detail in terms of data requirements. The tools differ
considerably as regards input data and the resources needed to run them, from a few man-
days for the riverine load apportionment model to several man-months for the fully dynamic
and distributed models such as NL-CAT and DAISY/MIKE-SHE. Moreover, the models
differ in their ability to be applied to scenario analysis, the simplest models being of limited
use for such purposes, whereas the models representing soil processes in a deterministic way
are very useful. The above mentioned nutrient source apportionment tools differ profoundly
in their approach to predict the diffuse nutrient losses from rural areas to surface waters. This
is caused by several factors: (i) their level of complexity; (ii) their representation of system
processes and pathways; (iii) resource (data and time) requirements. The quantification tools
range from complex, process-based models — which typically have demanding data
requirements — to semi-empirical (conceptual) meta-models with some export coefficients,
and approaches based on mineral balances and apportionment of the riverine load measured.
All source apportionment tools have strengths and weaknesses that should be taken into
consideration when choosing the most robust tool for a certain task.
Groundwater Interactions with Surface Waters 119

Sources of Pollution: Background Losses

The background loss of nitrogen is usually small compared with other diffuse and point
sources. Only in very sparsely populated areas is a major source. The background loss does
not reflect a reference condition, because most areas are subject to substantial atmospheric
depositions of ammonia and NOx from nearby or far away sources. The background loss of
phosphorus is relatively significant compared to other sources. It depends on the geological
conditions, and may differ even over short distances. In areas dominated by marine
sediments, there may be naturally high phosphorus concentrations in surface water and
groundwater.

Sources of Pollution: Agricultural Diffuse Losses

At large scale, agriculture is the single dominating source of nitrogen pollution, typically
contributing 50–80% of the total load. The situation may be different in smaller catchments
with high population densities (e.g. large cities), very poor wastewater treatment, or many
industrial facilities discharging poorly treated wastewater. Agriculture is also one of the
largest contributors to phosphorus pollution, along with various point sources. But contrary to
nitrogen, there are larger differences between the different regions and catchments. The
agricultural share is often about half the total load, and in most source apportionments it is
between 25 and 75%. The relative share is to a large extent affected by the point source share,
which is in turn a result of population density, industrial activities and wastewater treatment.

Sources of Pollution: Atmospheric Deposition

Both nitrogen and phosphorus are deposited in water and soil in different forms: nitrogen as
ammonia which has evaporated from animal manure, and as NOx coming from combustion of
fossil fuels, i.e. power plants and transportation; phosphorus as dust, falling leaves and bird
faeces. The deposition of ammonia nitrogen is of the same order of magnitude, and it is
highest in regions with high livestock densities. The proportion that falls on the ground is
usually not considered separately in source apportionments, but becomes a part of leaching
from the soil. The proportion that falls directly on the surface of inland or marine water is
assessed separately in several source apportionments. It is often very small compared with
other sources, but in lakes with a large surface area compared with the total catchment, or in
coastal or marine waters, it may constitute a significant part of the total inputs. The deposition
of phosphorus is generally small and difficult to estimate.

Sources of Pollution: Rural Population

The majority of the population in scattered dwellings is usually not connected to wastewater
treatment plants. The wastewater is discharged directly to surface water or to a percolation
system, possibly through a septic tank or other purification system. Scattered dwellings are in
principle point sources, but due to their abundance they are often considered as a diffuse
source or even as part of the agricultural contribution. At large scale, the annual discharges
120 A. Azzellino, R. Salvetti and R. Vismara

from scattered dwellings are typically 0.1–0.5 kg/ha for nitrogen and 0.01–0.1 kg/ha for
phosphorus, and may be occasionally higher (EEA, 2005). Scattered dwellings are relatively
insignificant sources of nitrogen and phosphorus. The contribution of nitrogen is generally
smaller than the background contribution and much smaller than the agricultural contribution.
The contribution of phosphorus is also small, usually of the same order of magnitude as the
background loss, but in some catchments may not be insignificant.

Sources of Pollution: Point Sources

In Europe, nutrient discharges from municipal wastewater treatment plants are in general
higher than for any other point source (EEA, 2005). Results from large inland and marine
catchments show that municipal wastewater constitutes about 75% of the point source
discharges of both nitrogen and phosphorus. Industrial sources constitute about 17% and
other point sources are also relatively insignificant. Locally, in smaller catchments, all types
of point sources may be significant in relation to pollution management.
Results from source apportionment studies are important in the policy formation process
and in monitoring the implementation of policies and the effectiveness of measures. In this
context, the source apportionment of nutrient loads should be carried out applying a relevant
source apportionment tool at regular intervals. The source apportionment could be done and
reported every three to five years for a representative part or for the entire network of river
stations. Data on nutrient loads and information on point source discharges, etc., can be used
for developing and/or calibrating models for diffuse nutrient losses. Models like the ones
described by Grimvall and Stålnacke, 1996; Kronvang et al., 1995; and Grizetti et al., 2005
are, in fact, as more effective and reliable as more data are available for the quantification of
annual nutrient discharges from the various point sources (i.e. sewage treatment plants,
industrial plants, scattered dwellings, fish farms, urban stormwater run-off, etc.) and as more
knowledge-based become the understanding of retention processes in watersheds. For this
reason, more robust data series are needed to quantify annual nutrient retention in streams,
lakes, reservoirs and riparian wetlands, to calculate average groundwater residence times in
the different hydrogeological regions and to evaluate the potential degradation of nitrogen in
groundwater aquifers.

Tools for Modeling Sources, Retention and Losses


The potential nitrogen and phosphorus inputs to primary surface water recipients are
transferred via a number of pathways. A large number of removal, storage or transformation
processes may influence the final quantities of nitrogen and phosphorus entering primary
surface water recipients. The nitrogen and phosphorus loss pathways to surface waters
include: losses by surface runoff (i.e. transport of dissolved nitrogen and phosphorus), losses
by soil erosion (i.e. transport of particular, adsorbed nitrogen and phosphorus), bank and
riverbed erosion, losses by artificial drainage flow (i.e. through drainage pipes/tile drainage),
losses by leaching (i.e. net mineralisation, percolating waters, interflow, tile drain flow, spring
water and groundwater), and direct atmospheric deposition on inland surface waters.
Groundwater Interactions with Surface Waters 121

The different loss processes and pathways are very complex and variable, the
significance of their effects also varies between nitrogen and phosphorus. It is therefore
inherently difficult to quantify diffuse losses accurately. In the absence of comprehensive
measurements, it is necessary to apply calculation methodologies (i.e. computer-based
modelling techniques). Using such tools, losses from diffuse sources of nitrogen and
phosphorus can be estimated either as the sum of all delivery pathways or as losses by every
individual pathway.
The losses of nitrogen and phosphorus via the various pathways can vary substantially,
depending on land use and management. In addressing the pathways, the following land cover
categories need often to be taken into account: agricultural land and categories such as
forests, unproductive land and unpaved urban areas not connected to sewerage.
Natural background losses of nitrogen and phosphorus constitute a part of the total
estimated nitrogen and phosphorus inputs to primary surface water recipients and include:
losses from unmanaged land; and that part of the losses of nitrogen and phosphorus from
managed land that would occur irrespectively of anthropogenic activities.
Direct atmospheric deposition of nitrogen on inland surface waters may represent an
important input and should be quantified where it is considered as a major source of the total
inputs of nitrogen to inland surface waters. The atmospheric deposition of nitrogen on land is
accounted for within the quantification of nitrogen and phosphorus reaching the primary
surface water recipients via the soil-related pathways. Atmospheric deposition of phosphorus
should be considered if the source is of significant importance, such as in areas where lakes
constitute a major part of the catchment.
In order to quantify the nitrogen and phosphorus reaching the primary surface water
recipients via the soil-groundwater system three levels can be considered:

level 1. net nitrogen and phosphorus inputs into the soil, e.g. mineral balances, loss
coefficients;
level 2. retention, sedimentation on land and nitrogen and phosphorus leaching from the
rootzone;
level 3. the total nitrogen and phosphorus inputs to the primary inland surface water
recipients via the various pathways.

Mineral balance calculations are appropriate for characterising the intensity of the
agricultural system, and can serve to indicate the likely magnitude of potential long-term
nitrogen and phosphorus losses, including ammonia volatilisation. Mineral balances are most
informative within regions of high nitrogen and phosphorus surplus, although diffuse nitrogen
and phosphorus losses are still possible even under zero or negative surplus conditions (e.g.
due to soil nitrogen mineralisation). Root zone losses of nitrogen may indicate the potential
input of nitrate to primary surface water recipients. Changes in agricultural management
practices will influence the nitrogen and phosphorus root zone losses, but they may not be
fully reflected in the figures of total nitrogen and phosphorus inputs to primary surface water
recipients, due to time lags caused by hydrological conditions and processes in the soil. It is
therefore helpful to quantify nitrogen and phosphorus losses from the root zone as this
intermediate level represents a valuable basis for both assessing the effects of nitrogen
reduction measures and for comparing different catchments.
122 A. Azzellino, R. Salvetti and R. Vismara

The potential inputs of phosphorus to primary inland surface water recipients are usually
much more difficult to quantify than the nitrogen inputs. In addition they are often more
spatially and temporally variable. Phosphorus losses are better correlated with soil P-
saturation, erosion risks and run-off pathways.
It is recognised that local environmental conditions and land management practices are
important factors regulating diffuse nitrogen and phosphorus losses to surface waters. The
choice of methodology is principally constrained by data availability and the understanding of
environmental processes at the scale of reporting. Thus, no single quantification methodology
can currently be recommended for any case of study.
Modeling can be combined with multivariate statistical techniques (e.g. principal
component or factor analysis) to improve our capability to “detect” the effect of the sub-
surface runoff on the water quality of specific water courses. With studies of this kind, it was
proved that the sub-surface runoff can be a significant source of diffuse pollutant even in dry
weather conditions.
Azzellino et al., 2006 showed that combining model simulations with a factor analysis
applied to instream water quality measurements, may help to identify potential hidden
features such as the existence of groundwater recharge effects. Factor Analysis, in fact,
outlining a lack of correlation between nitrates and nitrites, suggested a different source
apportionment for the two pollutants. On the other hand, the existence of a correlation
between nitrates and chlorides was interpreted as a groundwater fingerprint suggesting
groundwater as the common source of the two pollutants (see Table 1).

Table 1. Factor analysis of a lowland river (Mella river, Italy) where groundwater plays
a significant effect on the water quality of the river. Higher factor loadings are outlined
in bold

Factor loadings (VARIMAX rotation of Principal components)


1 2 3 4 5 6
D.O. -0.156 0.008 0.028 -0.005 0.937 -0.051
BOD5 0.048 0.442 0.004 -0.659 0.160 0.262
COD 0.072 0.388 0.560 0.389 -0.170 0.165
E.coli -0.102 0.212 -0.045 0.803 0.094 0.179
N - NH4 -0.028 0.831 -0.065 0.040 0.107 0.023
N - NO3 0.945 -0.039 0.048 -0.023 -0.073 0.059
P-tot 0.321 0.761 0.024 0.047 -0.222 -0.107
P - PO4 0.483 0.640 0.224 -0.025 -0.070 -0.117
N-tot 0.942 0.092 0.083 -0.016 -0.043 0.076
pH -0.526 -0.204 0.531 -0.169 0.210 -0.248
temperature 0.347 -0.082 0.808 -0.095 0.047 0.044
conductivity 0.907 0.191 0.098 -0.088 -0.067 0.042
hardness 0.904 0.200 0.107 -0.072 -0.046 0.059
SS 0.172 -0.130 0.040 0.052 -0.055 0.909
Cl - 0.876 0.264 0.070 -0.023 -0.125 0.105
SO4 2- 0.447 0.523 -0.216 -0.002 0.181 -0.152
% explained variance 32.5 16.0 8.7 8.1 7.0 6.8
% cumulative variance 32.5 48.5 57.2 65.2 72.2 79.0
Groundwater Interactions with Surface Waters 123

In such lowland rivers, Azzellino and colleagues (unpublished data) found that up to 60%
of the measured load could be apportioned as sub-surface-runoff-derived. In other case
studies, Factor Analysis did not outline a groundwater effect on the water quality since the
correlations found between pollutants suggested a different source apportionment (see Table
2). Singh et al. (2005) used principal component analysis (PCA), discriminant analysis (DA)
and partial least squares (PLS) in order to investigate the compositional differences between
surface and groundwater samples and the spatial variations in groundwater composition due
to the influence of natural and anthropogenic factors. In their study the groundwater samples
were generally dominated by contaminants deriving from both natural and anthropogenic
sources, whereas, surface waters were generally more influenced by contaminants of
anthropogenic/industrial origin.
Notwithstanding, to date, few studies have attempted to overlay measurements with the
conceptual modeling of the GW-SW system.
Some exemplary numerical experiments about the spatial variability and varying extent
of groundwater–surface water interaction have been accomplished by detailed two-
dimensional, vertical-plain simulations of exemplary transects of rivers beds connected with
the adjacent floodplains and lowlands by using numerical unsaturated-saturated soil water
models, e.g., HYDRUS-2D (Simunek et al., 1994; Joris and Fejen, 2003). However these
examples only refer to the field scale. At the mesoscale, several models have been developed
and applied for water balance simulations of lowland–floodplain landscapes, which account
for groundwater–surface water interactions in different manner. Some of these models are
based on more conceptual approaches for soil and groundwater, e.g., by considering data of
groundwater gradient or surface water distances (HECNAR, Jayatilaka and Gillham, 1996;
Jayatilaka et al., 1996), or on a combination of a vertical soil water simulation with a
simplified representation of groundwater flows towards the main drainage direction
(SWATRE, Spieksma and Schouwenaars, 1997), but also on fully three-dimensional
physically based numerical approaches for both soil moisture and groundwater
(VanderKwaak and Sudicky, 1999; Sudicky et al., 2000; Weng et al., 2003). Other
approaches couple physically based groundwater models (two-dimensional horizontal-plain,
or full three-dimensional) with a more conceptual model for the hydrological processes in the
unsaturated zone and at the soil surface (SWAT – MODFLOW Galbiati et al., 2006,
AGRIFLUX–MODFLOW Lasserre et al., 1999).
Most of the models mentioned above perform important simplifications: a) the spatial
extent of the interactions between groundwater and surface water is assumed to be constant,
or b) spatially and temporally transient changes of processes are considered in a rudimentary
manner only. However, the “most-physically” approach – a full three-dimensional numerical
solution of saturated-unsaturated zone processes (in both the soil and ground water domain) –
encounters limitations for mesoscale applications, because of its high computational demands
and problems with changing type or boundary conditions (e.g., varying saturation areas in
space and time), and insufficient knowledge concerning representative parameters of the soil
hydraulic properties for such models. The groundwater–surface water interactions between
floodplains and surface waters are represented in most of these models. However, the
influence of the areas, which are adjacent to the lowland–floodplain landscapes (i.e., their
hydrologic conditions at the non-surface-water boundary) are not taken into account by most
models (Hayashi et al., 1998; Krause and Bronstert, 2005, 2006).
124 A. Azzellino, R. Salvetti and R. Vismara

Table 2. Factor analysis of a different lowland river (Chiese river, Italy) where Factor
Analysis did not outline a groundwater effect on the water quality of the river. Higher
factor loadings are outlined in bold

Factor loadings (Varimax rotation of Principal components)


1 2 3 4 5 6
D.O. -0.302 -0.315 0.085 -0.020 0.185 0.785
BOD5 0.178 0.220 0.022 -0.020 0.844 0.012
COD -0.045 0.410 -0.176 -0.015 -0.373 0.711
E.coli 0.085 -0.777 -0.034 -0.004 -0.058 0.062
N - NH4 0.748 0.496 0.203 -0.014 0.212 -0.097
N - NO3 -0.081 -0.031 0.055 0.986 0.013 -0.007
P-tot -0.084 0.010 0.069 0.985 0.024 -0.003
P - PO4 0.719 0.472 0.172 -0.026 0.212 -0.096
N-tot 0.101 0.850 0.234 0.004 0.242 0.074
pH 0.167 0.292 0.901 0.060 0.007 -0.013
temperature 0.117 0.027 0.953 0.047 -0.031 -0.022
conductivity 0.915 -0.164 0.067 -0.069 0.068 -0.109
hardness 0.900 -0.264 -0.036 -0.073 0.032 -0.063
SS 0.321 0.298 -0.208 0.404 0.514 -0.125
Cl - 0.891 0.063 0.150 -0.009 0.131 -0.096
2-
SO4 0.512 -0.582 -0.409 -0.043 -0.209 0.023
% explained variance 25.5 17.0 13.3 13.3 8.6 7.4
% cumulative variance 25.5 42.5 55.8 69.1 77.7 85.1

Krause et al. (2007a) investigated water balance and groundwater dynamics of a


floodplain catchment in the Northeast German lowlands with consideration of the variable
interactions between the riparian groundwater and surface water.
In their study, the temporally and spatially variable exchange fluxes between
groundwater and surface water were showed by means of experimental and numerical
investigations. In the same study, a coupled soil water balance–groundwater model was used
for the quantification of exchange fluxes across the groundwater–surface water interface,
representative of a typical floodplain in Central European lowland catchments. Krause et al.
simulation results indicated substantial exchange fluxes between groundwater and the river,
which are subjected to intensive spatial and temporal variability. The intensities and also the
directions of exchange fluxes were characterised by transient alterations in particular stream
reaches. Groundwater–surface water interactions were found to control the groundwater
recharge dynamics in the floodplain and outweighed the influence of vertical percolation and
root water uptake.
So, as these authors clearly stated, considering the special ecological significance of low
flow seasons for rivers and floodplains, the high impact of groundwater–surface water
exchange fluxes during summer appears evident. During periods of extreme low flow
conditions, in fact, the fluxes are crucial in shaping the hydro-chemical conditions and may
result in ecological stress in periods of main vegetation growth or algal blooms. The results of
their study prove that, in lowland catchments, the eco-hydrological assessments cannot
neglect to consider the temporal and spatial dynamics of the groundwater–surface water
Groundwater Interactions with Surface Waters 125

interactions. Moreover, it should be stressed that, although groundwater contributions from


rivers of this kind may generally represent only 1% of the annual total discharge, during low
flow conditions the effect can be much higher, since around 30% of the runoff generated in
the catchment is originated by groundwater discharge from the riparian zone along river
stretches.
Using the same model Krause et al. (2007b) tested the impacts of changing land-use
conditions on the water balance of the floodplain. Simulating floodplain water balance for
four different land-use scenarios, these authors demonstrated how alteration of the landscape
cover effects changes in evapotranspiration and vertical groundwater recharge. It was
furthermore shown, how these changes in vertical groundwater recharge are counter-balanced
by both lateral groundwater flow, outlining once more the tight interactions between
groundwater and surface waters. These authors concluded that the alterations of land-use
characteristics cause only marginal effects on the overall groundwater recharge and
floodplain water balance.
Moreover, to analyse the impact of the design and structure of the drainage network,
Krause et al. (2007b) developed a scenario which assumed the total removal of the artificial
ditches of the floodplain, hence a dismantling of about 40% of the whole channel system. In
contrast to the effects of the land-use change, dismantling the channel network generated
significant deviations in the groundwater recharge dynamics. It was shown that the
dismantling of the artificial drainage channels in areas with shallow groundwater depths and
direct interactions with the channel network, the major effects are: a) less groundwater
recharge, and subsequently lower groundwater stages during the wet winter period, b) higher
groundwater stages caused by less groundwater discharge to the river in the dry summer
period. On the other hand, also in areas distant from the channel network and with a higher
groundwater depth, a significant effect was expected on the basis of the simulations: a long-
term increase of groundwater levels mostly due to the prevention of groundwater discharge
into the drainage channels in summer.
In general, scenario analysis emphasises the necessity of integrated approaches, which
combine land-use management as well as the layout of river/drainage channel network
(including the design of the cross sections), to reach the desired impacts and improvements of
the water balance dynamics of the floodplain.

Critical Aspects in Source Apportionment Studies: The Case of Scattered


Dwellings

Due to the different methodologies and approaches, results of source apportionment studies
are not always fully comparable. Differences in the estimation methods used for calculation
of the discharges and losses from sources, and the sources to be taken into account may
introduce bias between studies. Generally, the discharges from larger point sources such as
urban wastewater treatment plants and industries are estimated with a relatively high level of
confidence, and for source apportionments at national or large catchment levels these sources
account for the majority of the point source nutrient discharge. For such calculations, some
countries only include small wastewater treatment plants larger than 1,000 PE while others
may include small wastewater treatment plants down to 30 PE. If discharges from minor point
sources such as aquaculture and scattered dwellings/villages are taken into account, care
126 A. Azzellino, R. Salvetti and R. Vismara

should be taken when comparing the diffuse load from these studies with that from studies
where estimations are not made for these minor point sources. Developing an understanding
and modeling tools that can help quantify site-scale system processes and watershed-scale
cumulative effects of On-site Wastewater System (i.e. OWS) has been recently set as goal for
research.
This was exactly the goal of the National Decentralized Water Resources Capacity
Development Project (NDWRCDP, Siegrist et al., 2005). The project is the result of a
collaborative effort involving the Colorado School of Mines (CSM), Electric Power Research
Institute (EPRI), Systech Engineering, Inc., United States Geological Survey (USGS), and the
Summit County Environmental Health Department.
In this study the quantitative understanding of hydraulic and purification of the common
soil-based OWS processes has been improved at the single-system scale, and site-scale
models and decision-support tools have been developed and tested. Based on literature data,
cumulative frequency distributions were developed for concentrations of key characteristics
of domestic Septic Tank Effluent (STE) and for the transport/fate parameters that describe the
treatment of nutrients in soil and groundwater systems (McCray et al. 2005). Moreover,
Through a series of experimental studies, the understanding of bacteria and virus
transport/fate in soil-based OWS was enhanced (Van Cuyk, 2005). A numerical model
(referred to as the Biozone Algorithm) was set up to simulate a soil-based OWS and to enable
site-scale scenario analyses regarding hydraulic and purification performance (Beach and
Craigh, 2003). A major facet of this project involved the refinement, application, and testing
of an existing watershed-scale model, Watershed Analysis Risk Management Framework
(WARMF) for the Dillon Reservoir watershed in Summit County, Colorado (Weintraub et al.
2002). There are about 1,500 OWS in this watershed as well as other nonpoint and point
sources of pollution, and more than 600 onsite drinking water wells along with community
wells and surface water supplies. In addition, Dillon Reservoir provides 25% of the drinking
water supply for the City of Denver. In this project, WARMF was modified to include
explicit representation of OWS of different performance features, an integrated Biozone
Algorithm, and cumulative frequency distribution curves for source concentrations and
transport/fate parameters. The water quality effects of OWS were simulated using WARMF
and compared to a water quality dataset generated during the project. Simulations were also
completed to assess realistic decision-making scenarios concerning wastewater infrastructure
in the watershed and to determine the comparative effects on water quality. WARMF as well
as two other models, the Better Assessment Science Integrating Point and Nonpoint Sources
(BASINS)/Soil and Water Assessment Tool (SWAT) and Method for Assessment, Nutrient-
loading, and Geographic Evaluation (MANAGE) models, were set up, calibrated, and applied
to the Dillon Reservoir watershed. After setup and calibration, model simulations were
completed to examine current wastewater management scenarios and the simulation results
were compared to field monitoring data. Examination of future wastewater management
scenarios was also completed through watershed-scale model simulations (such as simulation
of the abandonment of OWS and connection of homes to a centralized wastewater treatment
plant [WWTP]).
As evidenced by source load mass balance calculations, WARMF and BASINS/SWAT
simulations, and water quality monitoring and analysis of spatial and temporal trends at the
watershed-scale, in the Dillon Reservoir watershed compared to urbanized development and
WWTP discharges OWS were not found a principal source of water pollutants. Based on
Groundwater Interactions with Surface Waters 127

WARMF simulations of different wastewater management scenarios, the alternative of


extending central sewers and the conversion of OWS to a central WWTP appears to offer
little or no benefit in terms of surface water quality protection, and in some cases may lead to
surface water quality degradation.
Despite the results of this specific and straightforward study, a better understanding about
water-quality impacts of OWS especially related to high-density development areas is still
needed. Many areas are served by OWS and drinking water wells. Information is also needed
about human-health and water-quality risks that might be created by development of OWS in
those areas. Although a few individual studies have been conducted in watersheds, the
cumulative non-point source phosphorous loading contributed by OWS is not clearly
understood nor agreed upon by local water-quality stakeholders. Also, little is clearly
understood about the actual contribution of failed OWS to phosphorous and other
contaminant loading in the watershed. The tools that are currently available are generally not
addressed to support decisions around use of OWS versus central sewer service or to compare
costs of various wastewater alternative solutions.

References
Agrawal, G.D., Lunkad, S.K., Malkhed, T., 1999. Diffuse agricultural nitrate pollution of
groundwaters in India. Water Sci. Technol. 29 (3), 67–75.
Azzellino A., L. Bonomo, R. Salvetti, and R. Vismara. 2006. Combined use of the EPA-
QUAL2E simulation model and factor analysis to assess the source apportionment of
point and non point loads of nutrients to surface waters. Sci. Total Environ. 371: 214–
222.
Beach, D. N. and J. E. McCray. 2003. “Numerical Modeling of Unsaturated Flow in Mature
Wastewater Soil Absorption Systems.” Ground Water Monitoring and Remediation.
23(2), 64– 72.
Bellos, D., Sawidis, T., Tsekos, I. 2004. Nutrient chemistry of River Pinios (Thessalia,
Greece). Environment International 30: 105– 115.
Campling, P., Terres, J.M., Vande Walle, S., Van Orshoven, J., Crouzet, P., 2005. Estimation
of nitrogen balances from agriculture for EU-15: spatialisation of estimates to River
basins using the CORINE Land Cover. Physics and Chemistry of the Earth 30, 25-34.
Cinnirella, S., Buttafuoco, G., Pirrone, N. 2005. Stochastic analysis to assess the spatial
distribution of groundwater nitrate concentrations in the Po catchment (Italy).
Environmental Pollution 133: 569–580.
Fan, A.M., Steinberg, V.E., 1996. Health implications of nitrate and nitrite in drinking water.
An update on methemoglobinemia occurrence and reproductive and developmental
toxicity. Regul. Toxicol. Pharmacol. 23, 35–43.
Forman, D., Al-Dabbagh, S., Doll, R., 1985. Nitrates, nitrites and gastric cancer in Great
Britain. Nature 313 (21), 620–625.
Fried, J.J., 1991. Sources of nitrates in fissure groundwater in the humid tropical zone—the
example of Ivory Coast. J. Hydrol. 113, 231–264.
EEA Report No 7/2005 - Source apportionment of nitrogen and phosphorus inputs into the
aquatic environment. EEA, Copenhagen 2005.
128 A. Azzellino, R. Salvetti and R. Vismara

Galbiati, L., Bouraoui, F., Elorza F.J., Bidoglio, G., 2006. Modeling diffuse pollution loading
into a Mediterranean Lagoon: Developement and application of an integrated surface-
subsurface model tool. Ecological Modelling 193, 4-18.
Gelberg, K.H., Church, L., Casey, G., London, M., Roerig, D.S., Boyd, J., Hill, M., 1999.
Nitrate levels in drinking water in rural New York State. Environ. Res. Sect. A 80, 30–40.
Goni I. B. 2006. Tracing stable isotope values from meteoric water to groundwater in the
southwestern part of the Chad basin. Hydrogeology Journal 14: 742–752.
Grimvall, A. and Stålnacke, P. 1996. Statistical methods for source apportionment of riverine
loads of pollutants. Environmetrics, 7, pp. 201–213.
Grizetti, B., Bouraoui, F., de Marsily, G. and Bidoglio, G. 2005. A statistical method for
source apportionment of riverine nitrogen loads. J. Hydrol., 304, pp. 302–315.
Guay B.E., .Eastoe C.J., Bassett R., Long A. 2006. Identifying sources of groundwater in the
lower Colorado River valley, USA, with δ18O, δD, and 3H: implications for river water
accounting. Hydrogeology Journal 14:146–158.
Gulis, G., Czompolyova, M., Cerhan, J.R., 2002. An ecologic study of nitrate in municipal
drinking water and cancer incidence in Trnava District, Slovakia. Environ. Res. Sect. A
88, 182–187.
Hayashi, M., van der Kamp, G., Rudolph, D.L., 1998. Water and solute transfer between a
prairie wetland and adjacent uplands. 1. Water balance. J. Hydrol. 207, 42–55.
Hudak, P.F., 2000. Regional trends in nitrate content of Texas groundwater. J. Hydrol. 228,
37–47.
Jayatilaka, C.J., Gillham, R.W., 1996. A deterministic-empirical model of the effect of the
capillary fringe on near-stream area runoff. 1. Description of the model. J. Hydrol. 184,
299–315.
Jayatilaka, C.J., Gillham, R.W., Blowes, D.W., Nathan, R.J., 1996. A deterministic-empirical
model of the effect of the capillary fringe on near-stream area runoff. 2. Testing and
application. J. Hydrol. 184, 317–336.
Joris, I., Fejen, J., 2003. Modeling water flow and seasonal soil moisture dynamics in an
alluvial groundwater-fed wetland. Hydrol. Earth Syst. Sci. 7 (1), 57–66.
Kattan Z. 2006. Characterization of surface water and groundwater in the Damascus Ghotta
basin: hydrochemical and environmental isotopes approaches. Environ. Geol. 51: 173–
201.
Knobeloch, L., Krenz, K., Anderson, H., Hovel, C., 1992. Methemoglobinemia in an infant—
Wisconsin. Morbidity Mortality Weekly Report 42 (12), 217–219.
Krause, S., Bronstert, A., 2005. An advanced approach for catchment delineation and water
balance modelling within wetlands and floodplains. Adv. Geosci. 5, 1–5.
Krause, S., Bronstert, A., 2006. Water balance simulations and groundwater–surface water
interactions in a mesoscale lowland river catchment. Hydrol. Processes, published online
24.05.07, doi:10.1002/hyp.6182.
Krause, S., Bronstert, A., Zehe, E. 2007a. Groundwater–surface water interactions in a North
German lowland floodplain – Implications for the river discharge dynamics and riparian
water balance. Journal of Hydrology (2007) 347, 404– 417.
Krause, S., Jacobs, J., Bronstert, A., 2007b. Modelling the impacts of land-use and drainage
density on the water balance of a lowland–floodplain landscape in northeast Germany.
Ecological modelling 200 (2007): 475–492.
Groundwater Interactions with Surface Waters 129

Kronvang, B., Grant, R., Larsen, S. E., Svendsen, L. M. and Kristensen, P. 1995. Non-point-
source nutrient losses to the aquatic environment in Denmark. Impact of agriculture.
Marine and Freshwater Research, 46, pp. 167–177.
LaBaugh JW, Winter TC, Swanson GA, Rosenberry DO, Nelson RD, Euliss NH Jr. 1996.
Changes in atmospheric circulation patterns affect mid-continent wetlands sensitive to
climate. Limnol Oceanogr 41: 864–870.
Lambs, L. 2000. Correlation of conductivity and stable isotope 18O for the assessment of
water origin in river system. Chem. Geol. 164: 161–170.
Lambs, L., Loudes, J.-P., Berthelot, M., 2002. The use of the stable oxygen isotope (O18) to
follow the water distribution and absorption in riparian woodlands. Nukleotika 47, S71–
S74.
Lambs L. 2004. Interactions between groundwater and surface water at river banks and the
confluence of rivers. Journal of Hydrology 288: 312–326.
de Lange, W.J. 1999. A Cauchy boundary condition for the lumped interaction between an
arbitrary number of surface waters and a regional aquifer. Journal of Hydrology 226
(1999) 250–261.
Lasserre, F., Razack, M., Banton, O., 1999. A GIS-linked model for the assessment of nitrate
contamination in groundwater. J. Hydrol. 224, 81–90.
Liu, G.D., Wu, W.L., Zhang, J. 2005. Regional differentiation of non-point source pollution
of agriculture-derived nitrate nitrogen in groundwater in northern China. Agriculture,
Ecosystems and Environment 107: 211–220
Mclay, C.D.A., Dragten, R., Sparling, G., Selvarajah, N., 2001. Predicting groundwater
nitrate concentrations in a region of mixed agricultural land use: a comparison of three
approaches. Environ. Pollut. 115, 191–204.
McCray, J. E., S. L. Kirkland, R. L. Siegrist, and G. D. Thyne. 2005. “Model Parameters for
Simulating Fate and Transport of On-Site Wastewater Nutrients.” GROUND WATER
Vol. 43, No. 4, July–August 2005: 628–639.
Van Cuyk, S., R.L. Siegrist, A. Logan, S. Masson, E. Fischer, and L. Figueroa. 2001.
Hydraulic and purification behaviors and their interactions during wastewater treatment
in soil infiltration systems. Water Research 35, no. 4: 953–964.
Mutch, J., 1998. The hydrologic cycle and water movement. Nitrate – agricultural sources and
fate in the environment –perspectives and direction. In: Proceedings of the workshop.
Eastern Canada Soil and Water Conservation Centre, pp. 3–7.
Nikolaidis, N.P., Heng, H., Semagin, R., Clausen, J.C., 1998. Nonlinear response of a mixed
land use watershed to nitrogen loading. Agric. Ecosyst. Environ. 67: 251–265.
Nolan, B.T., Ruddy, B.C., Hitt, K.J., Helsel, D.R., 1997. Risk of nitrate in groundwaters of
the United States—a national perspective. Environ. Sci. Technol. 31: 2229–2236.
Oenema, O., Roest, C.W.J., 1998. Nitrogen and phosphorus losses from agriculture into
surface waters; the effects of policies and measures in the Netherlands. Water Science
and Technology 37 (3): 19-30.
Oenema, O., Van Liere, L., Schoumans, O., 2005. Effects of lowering nitrogen and
phosphorus surpluses in agriculture on the quality of groundwater and surface water in
the Netherlands. Journal of Hydrology 304 (1-4): 289-301.
Peck, J.C. 2007. Groundwater Management in the High Plains Aquifer in the USA: Legal
Problems and Innovations. In The Agricultural Groundwater Revolution: Opportunities
and Threats to Development (M. Giordano and K.G. Villholth). CAB International 2007.
130 A. Azzellino, R. Salvetti and R. Vismara

Rozemeijer, J.C., Broers, H.P. 2007. The groundwater contribution to surface water
contamination in a region with intensive agricultural land use (Noord-Brabant, The
Netherlands). Environmental Pollution 148: 695-706.
Sear, D.A., Armitage, P.D., Dawson, F.H. 1999. Groundwater dominated rivers. Hydrol.
Processes 13, 255–276.
Shepherd, K.A., Ellis, P.A., Rivett M.O. 2006. Integrated understanding of urban land,
groundwater, baseflow and surface-water quality—The City of Birmingham, UK. Science
of the Total Environment 360 (2006): 180– 195.
Siegrist, R. L., J. McCray, L. Weintraub, C. Chen, J. Bagdol, P. Lemonds, S. Van Cuyk, K.
Lowe, R. Goldstein, and J. Rada. 2005. Quantifying Site-Scale Processes and Watershed-
Scale Cumulative Effects of Decentralized Wastewater Systems. Project No. WU-HT-00-
27. Prepared for the National Decentralized Water Resources Capacity Development
Project, Washington University, St. Louis, MO, by the Colorado School of Mines,
Golden, CO.
Simunek, J., Vogel, T., van Genuchten, M., 1994. The SWMS-2D Code for Simulating Water
Flow and Solute Transport in Two-Dimensional Variably Saturated Media Version 1.2
Research Report No. 132. Department of Agriculture, U.S. Salinity Laboratory,
Riverside, California.
Singh, B., Sekhon, G.S., 1978-9. Nitrate pollution of groundwater from farm use of nitrogen
fertilizers—a review. Agric. Environ. 4, 207–225.
Singh, B., Singh, Y., Sekhon, G.S., 1995. Fertilizer-N use efficiency and nitrate pollution of
groundwater in developing countries. J. Contam. Hydrol. 20: 167–184.
Singh K.P., Malik A., Singh V.K., Mohan D., Sinha S. 2005. Chemometric analysis of
groundwater quality data of alluvial aquifer of Gangetic plain, North India. Analytica
Chimica Acta 550 (2005) 82–91.
Schreiber, J. D., Rebich, R.A., Cooper, C.M. 2001. Dynamics of diffuse pollution from US
southern watersheds. Wat. Res. 35 (10): 2534–2542.
Schoumans, O. F. and Silgram, M. (eds), with contributions from Andersen, H. E. and
Kronvang, B. (2004). Review and literature evaluation of quantification tools for the
assessment of nutrient losses at catchment scale. Norwegian Institute for Water Research
(NIVA). Euroharp report 1–2003; NIVA report SNO 4739–2003, 120 pp.
Spalding, R.F., Exner, M.E., 1993. Occurrence of nitrate in groundwater—a review. J.
Environ. Qual. 22, 392–402.
Spieksma, J.F.M., Schouwenaars, J.M., 1997. A simple procedure to model water level
fluctuations in partially inundated wetlands. J. Hydrol. 196, 324–335.
Strebel, O., Duynisveld, W.H.M., Bottcher, J., 1989. Nitrate pollution of groundwater in
Western Europe. Agric. Ecosyst. Environ. 26, 189–214.
Sudicky, E., Jones, J., Brunner, D., McLaren, R., VanderKwaak, J., 2000. A fully-coupled
model of surface and subsurface water flow: model overview and application to the
Laurel Creek watershed. In: Bentley, L., Sykes, J., Brebbia, C., Gray, W., Pinder, G.
(Eds.), Proceedings of the XIII International Conference on Computational Methods in
Water Resources. Calgary, Alberta, July 26–29, 2000. A.A. Balkema, Rotterdam, pp.
1093–1099.
Tabacchi, E., Lambs, L., Guilloy, H., Planty-Tabacchi, A.M., Muller, E., Decamps, H. 2000.
Impact of riparian vegetation on hydrological processes. Hydrol. Processes 14, 2959–
2976.
Groundwater Interactions with Surface Waters 131

Townsend, M.A., Sleezer, R.O., Macko, S.A., 1996. Effects of agricultural practices and
vadose zone stratigraphy on nitrate concentration in ground water in Kansas. U.S.A.
Water Sci. Technol. 33 (4–5), 219–226.
Van Cuyk, S. M. and R. L. Siegrist. 2004. “Fate of Viruses in the Infiltrative Surface Zone of
Systems that Rely on Soil Treatment for Wastewater Renovation.” Proceedings of the
Tenth National Symposium on Individual and Small Community Sewage Systems.
American Society of Agricultural Engineers, Sacramento, CA. 387–399.
Van Loon, A.J., Botterweck, A.A., Goldbohm, R.A., Brants, H.A., van Klaveren, J.D., van
den Brandt, P.A., 1998. Intake of nitrate and nitrite and the risk of gastric cancer: a
prospective cohort study. Br. J. Cancer 78 (1), 129–135.
Van der Molen, D.T., Breeuwsma, A., Boers, P.C.M., 1998. Agricultural nutrient losses to
surface water in the Netherlands: impact, strategies, and perspectives. Journal of
Environmental Quality 27 (1): 4-11.
VanderKwaak, J.E., Sudicky, E., 1999. Application of a physically-based numerical model of
surface and subsurface water flow and solute transport. In: ModelCARE 99 Conference.
IAHS, 265, pp. 515–523.
Weintraub, L., C. W. Chen, W. Tsai, J. Herr, R. A. Goldstein, and R. L. Siegrist. 2002.
“Modifications of WARMF to Assess the Efficacy of Onsite Wastewater Systems on
Public Health.” WEFTEC Conference Proceedings. October 2002, Chicago, IL. Water
Environment Federation, Alexandria, VA.
Weintraub, L. H. Z., C. W. Chen, R. A. Goldstein, and R. L. Siegrist. 2004. “WARMF: A
Watershed Modeling Tool for Onsite Wastewater Systems.” Proceedings of the Tenth
National Symposium on Individual and Small Community Sewage Systems. American
Society of Agricultural Engineers, Sacramento, CA. 636–646.
Weng, P., Giraud, F., Fleury, P., Chevallier, C., 2003. Characterising and modeling
groundwater discharge in an agricultural wetland on the French Atlantic coast. Hydrol.
Earth Syst. Sci. 7, 33–42.
Winter T.C., Harvey J.W., Franke O.L., Alley W.M. 1998. Ground water and surface water –
a single resource. US Geol. Surv. Circular 1139.
Winter T.C. 1999. Relation of streams, lakes, and wetlands to groundwater flow systems.
Hydrogeology Journal (1999) 7:28–45.
Winter T.C., Rosenberry D.O. 1998. Hydrology of prairie pothole wetlands during drought
and deluge: A 17-year study of the Cottonwood Lake wetland complex in North Dakota
in the perspective of longer term measured and proxy hydrological records. Climatic
Change 40: 189–209.
Woessner, W.W. 2000. Stream and fluvial plain ground water interactions: rescaling
hydrogeologic thought. Ground Water 38, 423–429.
Wollschläger U., Ilmberger J., Isenbeck-Schröter M., Kreuzer A.M., von Rohden C., Roth K.
and Schäfer W. 2007. Coupling of groundwater and surface water at Lake
Willersinnweiher: Groundwater modeling and tracer studies. Aquat. Sci. Vol. 69, 2007.
Wood, F.L., Heathwaiteb, A.L., Haygartha, P.M. 2005. Evaluating diffuse and point
phosphorus contributions to river transfers at different scales in the Taw catchment,
Devon, UK. Journal of Hydrology 304:118–138.
Zhang,W.L., Tian, Z.X., Zhang, N., Li, X.Q., 1996. Nitrate pollution of groundwater in
northern China. Agric. Ecosyst. Environ. 59, 223–231.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 133-147 © 2008 Nova Science Publishers, Inc.

Chapter 4

MODELLING OF WATER-SOLID INTERACTIONS:


A DISCUSSION OF DIFFERENT APPROACHES

Marek Šváb1 and Lenka Wimmerová2


1
Institute of Chemical Technology in Prague, Technická 5, Prague 6, 166 28, Czech Rep.
2
Dekonta, a.s., Dřetovice 109, Stehelčeves, 273 42, Czech Rep.

Abstract
This contribution contains an explanation and discussion of two different approaches to
the modelling of water-solid interactions. The term ‘water-solid interaction’ means all
interactions that are important in the frame of remediation methods, groundwater treatment
and pollution migration, including adsorption, leaching of contaminants, minerals
weathering, etc.
The discussed approaches involve an exact modelling of interactions based on chemical
reactions as well as what is known as the ‘semi-empirical’ approach, which is applicable for
very complex systems. On one hand, the exact approach is very promising for a detailed
understanding of particular processes and can be specified by a possible transfer of results
among various systems. On the other hand, the semi-empirical approach can also be a simple
alternative for modelling very complex systems.
The calculation of the equilibrium solubility of zinc and copper in water under various
conditions (with and without ammonia as a complexation agent), including detailed speciation
of both metals complexes, is an example of the exact modelling. The influence of ammonia
nitrogen on the solubility of both metals is obvious: a local solubility maximum occurs at the
pH of 9.4. The results have been confirmed by equilibrium batch experiments with zinc and
copper hydroxides that have proven the local solubility maximum at the pH of 9.4.
The exact modelling is further demonstrated by the adsorption of an organic contaminant
from water on three adsorbents in a mixture demonstrating the distribution of the contaminant
between various adsorbents. Although these isotherms are not exact descriptions of particular
sorption reactions, the method of calculation demonstrates the principles of use of the exact
geochemical modelling for calculations of competitive sorption on different adsorbents.
As an example of the semi-empirical approach, a mathematical model aimed at predicting
the course of a continuous soil flushing process by use of the input data obtained from simple
batch laboratory experiments is described. An objective of the example is to apply this new
model to soil polluted by zinc and copper (11 949 mg/Kg-1 and 1 895 mg/Kg-1, respectively)
by flushing this soil with an ammonia nitrogen solution with the pH of 9.4. The model
predicts correctly the period of time needed for the removal of weakly-bound metal fractions,
134 Marek Šváb and Lenka Wimmerová

as well as estimates the overall removal efficiency of metals from the soil during the flushing
process.
Using the above-described examples, the main aspects of both approaches are
demonstrated. Finally, future research possibilities are discussed.

Keywords: modelling, exact approach, semi-empirical approach, water-solid interaction

Introduction
’Interaction between solid phases and aqueous solutions’ is a very general term for various
particular physical-chemical processes that are of great importance to a number of natural
sciences, as well as for technological and industrial processes (geochemistry, aquatic
chemistry, environmental chemistry, soil remediation, water treatment, mining, agriculture,
waste disposal, etc.). This article relates to interactions important in geochemistry,
groundwater treatment, soil remediation and other similar environmental issues where the
water-solid interactions strongly affect chemical species distribution in a particular system,
species migration through porous media, as well as changes in the system under different
surrounding conditions. If we consider an aqueous solution containing some dissolved species
in contact with a solid phase, the following processes can occur in the system: desorption of
the species from the solid to the liquid, adsorption of the species from the solution on the
surface of the solid, dissolution of precipitated species, precipitation of dissolved species,
changes of oxidation-reduction state of some species, and in some cases, also the formation of
new species through chemical/biochemical reactions and/or the species decomposition in the
same way. These processes cover a lot of particular reactions and other effects. In this way,
for example, we know about several mechanisms of adsorption (chemical sorption, physical
sorption, ion exchange, etc.) [1]. Such particular mechanisms in the pure and well-defined
systems have been studied in-depth by a number of researchers for many years. And it is
thanks to them that we have a good theoretical knowledge of these processes, as well as of
methods of how to quantify and calculate them; however, this theoretical knowledge is hard
to use with very complex systems where a lot of subsequent effects and their mutual
influences occur. Finally, when a real, natural system is considered for model description, we
must always count on some simplifications. This contribution explains the main approaches,
which are possible for the model description of various systems that differ in degrees of
exactness.

Approaches to Modelling
This chapter seeks to explain the basic aspects of several types of approaches to the modelling
of the water-solid interactions, depending on the degree of exactness considered in the model.
Since the border between the types of models mentioned in this chapter is not strict, it is a
question of the researcher’s subjective opinion to which category some particular model falls.
Nevertheless, it is always useful to consider the model type somebody is dealing with,
because this determines what we can expect from the model, as well as what the
transferability of the model to another system or to different conditions is.
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 135

Generally, if we want to carry out a model description of a particular system, we are


usually forced to assume several simplifications. This is the reason why we are forced to
assume some simplifications: the impossibility of obtaining the desired input data for the
exact description and/or the very complicated description of the system in which a lot of
various effects influencing each other are considered. A number of these simplifications
increase with the system’s complexity.
The modelling tasks can be divided into two main types – static and dynamic – according
to the physical arrangement of the system. The static systems are specified by permanent
contact between both phases (water/solid), while for dynamic systems a continuous flow of a
solution through solid particles is typical (generally named as porous media). As examples of
typical static tasks, we can list: the calculations of solubility of precipitates, minerals
weathering under various conditions, batch adsorption processes, and diffusion processes in
pores of solids. The dynamic systems involve, for example, the calculation of sorption filters,
soil flushing processes, and pollution migration with groundwater. Generally, the dynamic
systems are substantially more complex, and they usually work with similar descriptions of
the particular system. We must consider the additional effects in the dynamic state, such as
advection, dispersion, diffusion and other similar effects that influence the solution flow
through porous media. Due to the complexity of the dynamic systems, the models of such
processes are usually less exact than the static models, and they usually need a suitable form
of calibration. In following text, basic types of modelling approaches, from the point of view
of exactness, are discussed. Subjectively, we can divide these approaches into exact, semi-
empirical, and empirical.
The exact models can be specified by effort about detail and exact description of
processes considered in the system according to current state of knowledge. If possible, no
simplifications are assumed. Such models are usually applied in the simple and well-defined
systems, where the principals of basic mechanisms are known, including their quantification.
A characteristic of these models is that their input data are independent of a particular system
and its surrounding conditions, and thus, these data have general validity. Equilibrium
thermodynamic data (equilibrium constants, solubility products, etc.) can be named as an
example of such a data type. The character of the exact models provides their main
advantage. This consists of the result transferability among various systems. Therefore, the
exact approach is advantageous and promising if the desired input data are available. It can be
applied to various hydrochemistry or geochemistry tasks: calculation of solubility diagrams,
minerals weathering, redox and acidobasic equilibriums, probable precipitated solid phases,
adsorption on exactly defined adsorbents, etc. For such tasks, software solutions have been
already created, e.g. PhreeqC, the Geochemist’s Workbench, and the Visual Minteq. Though
the exact modelling can explain behavior and reaction mechanisms of aqueous systems very
well, it has its limit in input data, mainly in the case of real, natural systems (especially in the
presence of soil), in which the data can be hard or impossible to obtain.
For the semi-empirical modelling approach, it is typical that we want to quantify
particular processes running in the system, but without considering basic and detail
mechanisms of the system. The application of this approach needs experimental
determination of input parameters of the model, which is generally carried out directly with
the studied system. These parameters can be either correction factors of known equations
describing basic processes (in the pure systems), or specific parameters suggested for the
purposes of the model only, which covers more particular processes. The advantage of this
136 Marek Šváb and Lenka Wimmerová

approach is the fact that there is no limit in terms of complexity of the system to be modeled.
Simply, with increasing complexity of the system, the model semi-empirical parameters can
be modified in order to allow its experimental determination. The other advantage is the
possibility to model dynamic systems as well. The disadvantage of this approach consists of
the necessity to experimentally determine input data, which disables the transfer of the data to
other systems.
For a better explanation of the semi-empirical approach, let us consider the following
example. One of the widely used methods in water treatment is active carbon adsorption,
which can be used for removing various species from water, e.g. metal ions and organic
pollutants [2]. Various curves can be used for quantification of the relation between the
equilibrium concentration of the adsorbate in the solution and its sorbed amount. In most
cases, Freundlich or Langmuir isotherms are applied [3]; however, these isotherms do not
reflect particular sorption mechanisms and reactions that really occur on the sorbent surface.
Active carbon has a very complex structure, with a number of various active surface groups
that are responsible for its sorption properties. If we want to apply the exact approach, it
would be necessary to measure all particular reactions between these various groups and the
adsorbate. Since this task would be very complicated, we can decide not to consider all
particular reactions, but to measure only an equilibrium state in the particular system of active
carbon-pollutant and to characterize it through a suitable curve, e.g. Langmuir, Freundlich or
another mathematical function. The fact should also be considered that the Langmuir model
was prepared especially for the well-defined model adsorbents with one type of a uniform
surface reaction site, and thus its use for description of the sorption equilibrium on activated
carbon is groundless. However, the fact that it often provides good interpolation of measured
data is why the Langmuir curve is so popular. Although the model cannot be used for other
adsorbates and/or adsorbents, we can carry out the equilibrium calculations for various
concentrations, suggest a batch water-treatment system, make the first estimate of a suitable
dynamic system (a sorption filter), and at least partly understand the behavior of the system.
A similar approach represents a distribution coefficient for quantification of the
distribution of species between the solid and liquid phase, which is frequently used in the
description of interactions between soil and groundwater. From the point of view of the solid
phase, soil represents one of the most complex matrixes containing both various types of
inorganic components (clays, sand, silt, etc.) and of organic matter (fulvic acids, humic acid,
etc.). Soil composition results in a number of various possible surface interactions and
reactions, which are almost impossible to define. Furthermore, in the case of toxic metals, a
number of different forms in which the metals can be present in soil are known. Finally,
although the distribution coefficient does not reflect the mechanisms of the interactions
between the phases, it provides relatively reliable data for the particular soil and groundwater.
The third type of modelling approach of the water-solid interactions is the empirical one.
It is typical for this approach that we accept our studied system as a ‘black box’. The
empirical approach needs experimental measuring of the system, observation of its reaction
on modified surrounding conditions, and interpolation of measured data by a suitable
empirical equation. The advantage of the empirical model consists in possibly dealing with
the complex and dynamic systems, although transferability of data is very limited. The data
and results are valid strictly for the same conditions and the range of modified parameters as
those used in the experimental measuring.
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 137

As an example of the empirical approach, determination of the width of an adsorption


zone (WAZ) in a sorption filter can be given [2]. WAZ, the important parameter of the
sorption process in the dynamic arrangement, depends on many complex parameters, e.g.
diffusion in adsorbent pores, the manner of the solution flow through a layer of adsobent, and
the flow rate. The WAZ can be experimentally determined in a column through measuring
concentration of specie in the column effluent till then its concentration is equal to its
concentration in the column input. If we are interested in knowing how WAZ depends on the
solution linear flow rate through the adsorbent layer, we have to measure WAZ for several
flow rates and make an interpolation. We can use the obtained dependence for optimization of
flow rate through the sorption filter, although we cannot simply transfer the results to the full-
scale design of the sorption filter due to its different geometry and conditions. Thus,
subsequent verification of the technological process is necessary. Furthermore, it is a question
whether the empirical approach should be named as ‘modelling’ because generally it means
only interpolation of experimentally observed reactions of the system in various conditions.
The above-discussed exact and semi-empirical modelling approaches to the water-solid
interactions are further demonstrated on three examples.

Example 1
The Example 1 demonstrates the exact modelling approach. Solubility of zinc and copper
from their hydroxides (or oxides) in an ammonia nitrogen solution can serve as a good
demonstration. It is a static system and input data for the model are the tabled thermodynamic
equilibrium data (i.e. solubility products, complexes stability constants, ammonia dissociation
constant). Thus, such data have general validity and are independent on the modeled system.
We can find it in thermodynamic tables or in various databases.
It is known that ammonia creates strong complexes with zinc and copper. The knowledge
of zinc and copper behavior in the presence of ammonia nitrogen is important when
considering for example polluted soil. Theoretically, ammonia as the complex agent may
provide selective increase of solubility of both metals in neutral pH, which would possibly
leach these metals from soil. It is the good example of simplification we have to accept: we
have to consider both metals in hydroxides’ (oxides’) forms, although we know that both
metals are presented also in many other forms in soil. For the exact calculation of metals
solubility, we are forced to assume it to be able to obtain input data required for the
calculation.
Let’s explain the modelling task. We want to know, what is the thermodynamically stable
state of the system containing Zn2+, Cu2+, NH3 (at defined concentration expressed as
concentration of ammonia nitrogen) and water under various pH. Particular reactions
considered in the system involve dissociation of water and ammonia, dissolution of both
hydroxides and formation of hydroxo- and amino-complexes of both metals. These simple
reactions were already presented including several results in our earlier study [4]. But the
calculations were carried out manually with limited accuracy and without consideration of
activity coefficients different from one.
At the present study, we used the commercial software for the exact hydro-geochemical
modelling ‘the Geochemist’s Workbench’ (GWB) (module ‘React’) with its basic
thermodynamic database [5]. We extended this dataset only about several complexes of
138 Marek Šváb and Lenka Wimmerová

copper with OH- and NH3 ligands because not all were included originally. If used correctly,
the software provides accurate results in relatively simple way.
The first result of the GWB software we would like to present is shown on theREF Figure
1. Total amount of zinc in the system was substantially higher than the soluble species
marked in the graph. Thus, the solid phase of mineral Zincite (ZnO), as the
thermodynamically most stable mineral under considered conditions, exists in the system
above the sum of all curves. It is obvious from the REFFigure 1 that if no ammonia is present
in the solution, the solubility of zinc between the pH 8.5 and 12 is extremely low (which is
commonly known) because the ammino-complexes are responsible for soluble species of zinc
within this pH range. Therefore, we find the local solubility maximum of zinc in the pH about
9.4-9.5, in which the solubility of other metals (iron, calcium, etc.) is very low. We can obtain
similar diagrams (as shown on the Figure 1REF) for copper, various ammonia concentrations
as well as for the system containing both metals together, where their behavior would be
quantitatively different thanks to a competitive formation of ammino-complexes of both
metals. Finally, such exact diagram obtained by use of the thermodynamic data provides great
opportunity to understand mechanisms and particular reactions running in the studied system.

Figure 1. Distribution of zinc-containing species in the solution of 0.05mol/L ammonia nitrogen under
various pH.

For verification of the model calculation of solubility diagrams, the equilibrium leaching
experiments were carried out with zinc and copper hydroxides. Metal hydroxides were
prepared by neutralization of zinc and copper salts to the final pH of approximately 9.5.
Solubility of zinc and copper were observed at the pH 8-10.5 and for ammonia nitrogen
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 139

concentrations of 0.2 and 0.3 mol/L. A methodology of the experiments is described in our
earlier study [4]. The Figure 2REF discloses the comparison of model and experimental
results. It is obvious from the graph, that agreement between the model and experimental data
is excellent, especially when considering the fact that the model results were obtained by use
of commonly valid data from the thermodynamic tables and the database released with GWB.
Looking to our earlier study [4], it is also obvious that fit between the model and the
experiments is better in case of GWB results (Figure 2REF) than of the ‘hand-made’
calculations used before. The reason probably consists in not considering activities different
from concentrations in our earlier study while GWB works with the activities. The Figure
2REF confirms that the local solubility maximum of zinc in the ammonia nitrogen solution
(pH 9.4) really exists. Similar result may be obtained for copper.

0.9 0.2 M
0.8 0.3 M
0.7
C(Zn) (g/L)

0.6

0.5

0.4

0.3

0.2

0.1

0
7 8 9 10 11 12
pH

Figure 2. Solubility of zinc from its oxide (Zincite) in the ammonia nitrogen solution: comparison of
the model results and the experimental data.

Since we know about the local solubility maximum of zinc and copper in the ammonia
solution, we can assume similar behavior of both metals for their different solid forms. For
example, if we consider ZnS as the solid form, we will obtain similar (but lower) solubility
maximum at the pH of approximately 10. This is caused by extremely low solubility products
of sulfides (lower than these of oxides).
We can also assume similar behavior of both metals if they are sorbed on various
surfaces. However, the exact modelling is impossible in this case because necessary
thermodynamic data are not available. Theoretically, we can measure the equilibrium sorption
data and use them in the calculation. Furthermore, we cannot correctly speak about the exact
modelling in this case as it will be explained in the Example 3.
To conclude, the exact modelling example indicated the local solubility maximum of
both metals in the ammonia solution and thus, ammonia might be usable for both metals’
140 Marek Šváb and Lenka Wimmerová

leaching from soil under such conditions. The existence of the local solubility maximum was
qualitatively confirmed also in polluted soil [4]. However, the exact modelling cannot be used
for the quantitative prediction of metals behavior in soil. We are not able to obtain required
exact thermodynamic data because it would be necessary to exactly describe all particular
interactions of metals with the soil matrix separately. Finally, although we have software
instruments for exact calculations (e.g. GWB), we are limited by the desired input data that
are impossible to measure in many cases. The exact modelling is a great instrument for
understanding of mechanisms in the relatively simple and defined systems but it could not be
used in the systems where we are not able to obtain the exact thermodynamic input data. This
is why, we have to accept definite degree of empirical measuring and thus the semi-empirical
model application in case of polluted soil.

Example 2
The Example 2 demonstrates modelling of organic specie adsorption on two adsorbents
(active carbon) in the solution. The example shows the approach which is falling somewhere
between the exact and semi-empirical methods.
The exact geochemical software of GWB is used for the calculations through the
Langmuir isotherms (the ‘exact’ part of the Example 2) but sorption mechanisms and
particular sorption reactions between complex active carbon surface and the adsorbed organic
specie are not known (the empirical part – we are aware that sorption occurs but we do not
know the particular mechanisms responsible for it). Furthermore, in principle there is no
serious reason for the application of the Langmuir isotherm for description of the equilibrium
between active carbon and organic specie. The reason for that is the fact that this isotherm
was deduced for the well-defined model cases where only one type of a uniform sorption site
is presented on the adsorbent surface and it reacts with the adsorbed specie with the 1:1 ratio
[6]. Although it is clear that such conditions are very far from the complex adsorbent like
active carbon, the Langmuir isotherm often provides good interpolation of measured
equilibrium data. This is why (together with its relative simplicity) this isotherm is so popular
for sorption equilibrium descriptions. But in principle we can interpolate measured
equilibrium sorption data by any suitable mathematical function.
Thus, we can conclude that the application of the Langmuir isotherm for active carbon
adsorption is the application of the exact model in the semi-empirical way. Especially, the
number of adsorption sites (presented in the Langmuir model) must be understood strictly
only as the empirical mathematical parameter, which has nothing to do with adsorption
mechanisms or active carbon surface characterization. It expresses only what the approximate
sorption capacity of active carbon towards particular specie is.
In this demonstration, naphthalene is used as a good example of organic pollutant and
two types of granulated active carbon will serve as the adsorbents. The Silcarbon Co.
(Kostelec nad Labem, Czech Republic) supplies both samples of active carbon (trade names
K835 and S835). The used samples have particle size between 0.5-2.5 mm, are steam-
activated and have similar surface area about 850-900 m2/g. The difference of these carbons
consists in used raw material; K835 is made from coconut shells while S835 from bituminous
coal. This is why the K835 sample has a higher content of micropores.
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 141

The input data needed for the model description (i.e. Langmuir isotherm parameters)
were determined experimentally. It is obvious that such data are valid only for the particular
active carbon and naphthalene but, for example, it allows calculations of different
concentrations. Thus transferability of the input data is limited on the particular system only.
The equilibrium sorption batch experiments were carried out with various doses of active
carbon (both types separately) to various volume of the naphthalene aqueous solution (initial
naphthalene concentration was limited by naphthalene solubility). The mixtures were shaken
for 24 hours to achieve the equilibrium. After it, the samples were filtered and naphthalene
residual concentration in the solution was determined by the UV spectrophotometric method
[7]. The obtained sorption isotherms including its interpolation by the Langmuir isotherms are
shown on the Figure 3REF.

40 35
S835 K835
30

C (adsorbed) (mg/g)
C (adsorbed) (mg/g)

30
25
20
20
15

10 10
C (adsorbed, max) = 37.5 mg/g C (adsorbed, max) = 29.7 mg/g
adsorption coefficient = 0.964 (L/mg) 5 adsorption coefficient = 1.690 (L/mg)
0 0
0 5 10 15 0 5 10 15
C (solution) (mg/L) C (solution) (mg/L)

Figure 3. The measured equilibrium sorption isotherms of naphthalene from the aqueous solution on
two types of active carbon (K835 and S835).

From the Figure 3REF it is obvious that both types of active carbon have high sorption
abilities towards naphthalene. The Langmuir interpolation does not seem to be perfect but the
experiments were repeated with the similar results several times. The deviations are probably
caused by the system complexity, low initial concentrations of naphthalene in the solution and
the analytical procedures consisting of vacuum filtration, extraction and measuring on a
spectrophotometer. Anyway, the results are suitable for the demonstration of the model
calculation. It was found that the K835 active carbon has higher maximal sorption capacity
(i.e. ‘number of sorption sites’, see above) towards naphthalene, what may reflect higher
content of micropores on its surface. On contrary, the adsorption coefficient (representing the
equilibrium constant of the sorption reaction) is lower than in case of S835, what means that
naphthalene is bound on its surface with less strength. We cannot suggest more from obtained
results, since we have no information about detail composition of their surface, particular
surface groups and their interactions with naphthalene. Furthermore, such data are valid
strictly for this system and for naphthalene as the only one sorbate.
The exact GWB software can be also used for the demonstration of the adsorption
calculation. For this purpose, the thermodynamic dataset of sorption reactions must be
prepared. The desired data are the equilibrium constant of adsorption reaction (i.e. adsorption
coefficient) and the number of adsorption sites. Both parameters were determined in the
laboratory experiments (see Figure 3REF). The sorption reaction discussed can be written as
(similarly for S835):
142 Marek Šváb and Lenka Wimmerová

K835: + naphthalene (aq) = K835:naphthalane (1).

The symbol ‘K835:’ in the equation (1) represents the hypothetical sorption site on
surface of active carbon K835, while ‘K835:naphthalene’ means surface specie on the active
carbon surface after adsorption of naphthalene. Although the equilibrium constant of this
reaction as well as number of adsorption sites on the surface (i.e. initial concentration of
K835:) was measured, it is obvious that such a simple reaction (according to the Langmuir
model) cannot precisely describe the adsorption on very complex surface of active carbon
with number of various groups. This is why the use of the Langmuir isotherm represents
substantial simplification of the system. Furthermore, although the Langmuir isotherm is the
exact model of adsorption, it is applied semi-empirically after simplification of the complex
system (as the active carbon is) to the form of equation (1). The presented description of the
sorption equilibrium allows calculation of various tasks including the equilibrium
composition of the system containing both adsorbent and naphthalene, number of batch
adsorption steps needed for achieving desired residual concentrations of naphthalene in water,
etc. Solution of such tasks has been already described in the literature [2].
In this example, the calculation of the equilibrium composition of the mixture containing
together both types of active carbon is demonstrated by use of the GWB software. The
example is useful for understanding of basic rules important in equilibrium systems and for
demonstration of the calculation of competitive adsorption on two adsorbents.

Figure 4. The equilibrium composition of the system containing naphthalene and active carbon K835
(0.25 g) into which up to 0.8 g of active carbon S835 is added.
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 143

Let’s assume the following system: aqueous solution (1 L) of naphthalene (initial


concentration 25 mg/L) containing active carbon K835 (0.25 g) to which 0.8 g of active
carbon S835 is added stepwise. The GWB provides the equilibrium composition of the
system in initial and any other states (i.e. after addition of different amounts of S835 till final
value of 0.8 g). The results are shown on the Figure 4REF.
The initial composition of the system is obvious from y-axis when the addition of active
carbon S835 is zero. In this situation, the surface of K835 is mostly occupied by naphthalene,
since the equilibrium concentration of free adsorbing sites (K835:) is very low. The high
adsorption degree of naphthalene also corresponds with its high residual concentration in the
solution. It means that the equilibrium of reaction (1) is moved towards its product. When
second adsorbent (S835) starts to be added, the naphthalene is adsorbed mostly on its surface
from the solution but not through desorption from the K835 surface (till addition about 0.4 g).
This observation would be hard to obtain without model calculation, because it reflects the
differences between the adsorption coefficients and it is influenced by mass of all media and
naphthalene as well. After the addition of S835 about 0.4 g the naphthalene starts to be
desorbed from K835, however it needs both relatively a high dose of S835 what results in
higher equilibrium concentration of adsorbing sites of S835 (S835:) and lower residual
concentration of naphthalene in the solution. From the S835 added amount about 0.7 g,
further addition of this active carbon causes low decrease of naphthalene concentration in the
solution only as well as low increase of its amount adsorbed on the S835 surface
(S835:naphthalene). Thus, further addition of S835 is ineffective (from the point of view of
naphthalene residual concentration in the solution and degree of exploitation of S835
adsorption capacity). It is obvious that in this case bound of naphthalene on K835 is so strong
so even the high dose of S835 does not cause substantial desorption of naphthalene from
K835.
Thus, the REFFigure 4 provides the good example of quantitative understanding of
behavior of the systems containing more than one adsorbent. It would be possible to calculate
other similar tasks by GWB, but it is necessary to remember that the data are valid for the
particular system only. This is a key difference from the Example 1, where the data have
general validity. Although GWB is the exact modelling system, it may be used successfully
for design and/or optimization of the adsorption technology as well.

Example 3
The last example demonstrates the semi-empirical approach to modelling of soil flushing
process. In contrast to the Example 2, in which the simplified but the exact mechanism of
sorption was applied (i.e. the Langmuir isotherm), the Example 3 shows strictly the semi-
empirical approach that has no exact physical-chemical interpretation. This method of the
model description allows calculation of such complex system as soil is.
Metals-polluted soil may serve as a good example. As it was mentioned before, the main
problem disabling the exact modelling consists in soil complexity given by its composition as
well as fractionation of metals onto several forms. Following fractions of metals are usually
considered: soil solution, exchangeable/carbonates, reducible – adsorbed on iron/manganese
oxides, oxidizable – adsorbed on organic matter/bound in sulfides and residual fraction [8].
Complex soil composition (various minerals, clays, sand, dust and non-defined organic matter
144 Marek Šváb and Lenka Wimmerová

with its adsorbing surfaces, etc.) is also important [9]. If the exact modelling is applied, it is
necessary to know all thermodynamic data for all species and forms of metals presented in
soil. Such data would be then independent on a particular system and a modeler would be
able to choose only the system composition (similarly as in the Example 1). It is obvious that
in case of polluted soil obtaining such exact data is almost impossible. Theoretically, if the
demanded exact (system-independent) data are available, the exact modelling is possible.
Nevertheless, it is a question of many future years when such detail thermodynamic data for
soil components and pollutants will be available. Thus, at present state of art only application
of the semi-empirical approach remains.
Following example demonstrates the model of leaching of zinc and copper from
contaminated soil by the ammonia nitrogen solution. It was already shown in the Example 1
that the local solubility maximum of both metals exists at the pH of 9.4 when
oxides/hydroxides are considered as the solid phases. Although it is clear that soil
composition is substantially more complex than only oxides/hydroxides as the solid forms of
zinc and copper, it can be assumed that ammonia will have definite similar influence also on
different forms of metals present in soil. However, it is impossible to exactly understand the
reactions mechanisms as well as to differentiate contribution of each form to total solubility
of both metals. It is good opportunity to understand possible procedures if the modelling of
the soil-solution interactions is needed. Firstly, the exact model (GWB) can be applied in
order to get to know the basic principles and behavior of the system in the model conditions
through the system-independent thermodynamic input data. Secondly, knowledge from the
exact part can be applied semi-empirically in a complex matrix (containing the same
pollutants) where the exact approach is impossible (i.e. in soil).
The Example 3 extends our previous published study in which a new semi-empirical
model of soil flushing was explained in detail [10]. For detail understanding of this model it is
recommended to read through this article because in this contribution only basic principles
will be discussed in order to explain differences from two above-described examples.
The semi-empirical description of interaction liquid-solid phase (i.e. soil in our case) can
be specified by its quantification through some suitable parameter. The parameter known as
‘distribution coefficient’ (further DC) is very popular and is widely used for this purpose. It is
defined easily as a ratio of metal equilibrium concentrations in a solution and soil (or
reversibly). Although the DC is exactly defined, it is the semi-empirical parameter only
because it says nothing about a reaction type and/or a mechanism, phase composition, etc.
This coefficient expresses only what the equilibrium distribution of a metal among phases is.
DC can be easily determined experimentally and thus it allows calculations of the equilibrium
distribution of metals in the studied system (only in which DC was determined). However,
such approach is not acceptable in polluted soil, where a number of various forms of metals
are presented with substantially different leachability. This situation results in non-constant
but a concentration-depended value of DC, which reflects the various forms of metals
presented in soil. Finally, in case of polluted soil, the concentration dependence of DC has to
be measured considering the various forms in which the metals are presented in the soil.
The concentration dependence of DC can be measured by a set of equilibrium batch
experiments carried out with the particular soil and a leaching solution under various
solution/soil ratios. In our example, the leaching media was the ammonia nitrogen solution
(pH 9.4) as in case of the exact modelling (see Example 1). The soil properties and
experiments methodology are described in our previous study, where also the results achieved
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 145

for copper are presented [10]. At present paper, the results related to zinc are demonstrated
(seeREF Figure 5).

50
20 mL 0.1 M
40 0.2 M
0.3 M
Kws (Zn) (g/L)

0.5 M
30 50 mL 0.6 M
0.8 M
20 100 mL

250 mL
10

500 mL
0
4000 6000 8000 10000 12000
Cs* (Zn) (mg/Kgdry soil)

Figure 5. The measured zinc distribution coefficients for various ammonia concentrations and volumes
of the solution with the same dose of the soil (2 g) under the pH value of 9.4. (Cs* - equilibrium zinc
concentration in the soil, Kws-distribution coefficient solution/soil, unit in the figure: gdry soil L-1solution)

From the REF Figure 5 it is obvious that DC of zinc depends strongly on both ammonia
concentration in the leaching solution and the residual equilibrium concentration of zinc in
soil. The different residual concentrations of the metal in soil were achieved for each
ammonia concentration by changing the solution volume under the same dose of soil. Thus,
the demonstrated concentration dependence of the metal DC proves that the assumption of the
constant distribution coefficient (commonly used in the models previously suggested) means
an excessive simplification. The REF Figure 5 also expresses what the content of ‘easily
leachable’ fractions of zinc in soil approximately is. However, it says nothing about exact
mechanism and/or about a fraction mainly responsible for this leachability. Nevertheless,
from the practical point of view, obtaining such information is very useful.
The REF Figure 5 describes the equilibrium distribution of zinc between solution and
soil, similarly as for example sorption isotherms in case of adsorption processes. There is no
effort to at least simplified understanding of exact reaction mechanisms running in the
system. It is known only, that amino-complexes are responsible for the zinc solubility at the
pH of 9.4. Despite to this unknown mechanism, the results allow to make the calculations of
the equilibrium distribution of zinc even in such complex system. Furthermore, it also allows
calculations of a continuous process of soil flushing by use of special software that uses the
interpolated results from the REF Figure 5 as the input data [10]. Again, it is important to
remember that such data are valid for the particular soil and solution only, for which the DC
dependence was measured.
146 Marek Šváb and Lenka Wimmerová

Conclusion
This contribution is focused on demonstration of various modelling approaches of the water-
solid interactions. The presented approaches are further elucidated by three practical
examples.
The exact modelling is very useful for detailed understanding of mechanisms running in
the system. Usually, its input data are independent on the particular system. Thus, it can be
transferred to other conditions. The limits of this approach consist in unavailability of the
desired input data for more complex systems (adsorption processes, soil, etc.).
The semi-empirical approach provides the great advantage since it can be also used in the
very complex systems. Such approach can be further divided on two cases. The first case
reflects effort to apply exactly deduced rules and/or equations in the simplified way. As
typical example, the application of the Langmuir isotherm for description of adsorption in the
complex systems (i.e. active carbon) can be served. The second case consists in the
application of the suitable semi-empirical parameter for description of the equilibrium
distribution (e.g. in polluted soil) without any effort to understand the interaction
mechanisms.
Thus, the main conclusion coming from this contribution is that at modelling of real
technological and/or natural systems, the definite degree of both simplification and the semi-
empirical approach must be accepted mainly because of unavailability of desired input data
for the exact modelling. It is necessary to remember all simplifications during interpretation
or other further use of model results as well as to be aware that the model result quality
reflects reliability of the input data. Hence, the modelling is only the instrument for further
understanding of the system. Model users should be always aware that the modelling can
provide completely wrong results (or the results that may be wrongly interpreted) if they do
not understand the basis of the model.
Generally, it can be expected that the future research in the modelling field will be more
and more focused on the exact approach, which is promising for detail understanding of the
problem as well as for a better practical exploitation of the modelling results. This is why the
demand for the serious and precise thermodynamic data will be increasing because the
absence of the desired data represents the reason why the exact modelling cannot be used in
more cases. On contrary, it can be also expected that number of tasks that will need to apply
the semi-empirical approach remains because generally valid exact input data will not be
available. Time will show where will be a new border between the exact approach and the
semi-empirical modelling as well as which tasks currently modeled by the semi-empirical
methods we would be able to solve in the exact way. Thus, moving this modelling border as
far as possible represents a big challenge for researchers in future.

Acknowledgement

This contribution was co-financed by the Czech Ministry of Industry and Trade (grant no. FT-
TA3/077).
Modelling of Water-Solid Interactions: A Discussion of Different Approaches 147

References
[1] Pitter, P. Hydrochemie, 3rd edition, Institute of Chemical Technology in Prague, Prague,
Czech Republic, 1999, pp 54-55.
[2] Sincero, A. P., Sincero, G. A. Physical-Chemical Treatment of Water and Wastewater,
CRC Press, Boca Raton, Florida, USA, 2002.
[3] Yang, R. T. Adsorbents: Fundaments and Application, J.Wiley&Sons, Hoboken, New
Jersey, USA, 2003.
[4] Kubal, M., Svab, M., Borysek, A., Cermak, J. Separ Sci Technol 2001, 36, 3223-3238.
[5] Bethke, C. M. Geochemical Reaction Modelling, Oxford University Press, New York,
USA, 1996.
[6] Voleski, B. Sorption and Biosorption, BV Sorbex inc., Montreal, Canada, 2003.
[7] Jaffé, H. H., Orchin, M. Theory and Applications of Ultraviolet Spectroscopy,
J.Wiley&Sons, New York, USA, 1964.
[8] Zemberyova, M., Bartekova, J., Hagarova, I. Talanta 2006, 70, 973-978.
[9] van Loon, G. W., Duffy, S. J. Environmental Chemistry: A Global Perspective, 2nd
edition, Oxford University Press, Oxford, UK, 2005.
[10] Svab, M., Zilka, M., Müllerova, M., Koci, V., Müller, M. Sci Total Environ 2008, 392
(2-3), 187-097.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 149-166 © 2008 Nova Science Publishers, Inc.

Chapter 5

FUNDAMENTALS OF GROUNDWATER MODELLING

Husam Baalousha*
Hawke’s Bay Regional Council, Private Bag 6006, Napier, New Zealand

Abstract
Groundwater Modelling is an efficient tool for groundwater management and
remediation. Models are a simplification of reality to investigate certain phenomena or to
predict future behaviour. The challenge is to simplify reality in a way that does not adversely
influence the accuracy and ability of the model output to meet the intended objectives.
Despite their efficiency, models can be complicated and produce wrong results if they are
not properly designed and interpreted. Regardless of the type of model being used, similar
sequences should be followed in modelling. To help selecting the proper model, modelling
objectives should be clear and well identified.
If the conceptual model is not properly designed, all modelling processes will be a waste
of time and effort. To build a proper conceptual model, hydrogeological data should be
sufficient and reliable. Calibration and verification are the last steps in modelling before
writing the final model report.
This chapter discusses the stepwise methodology of groundwater modelling with
explanation of each step. It contains a brief description of different types of models and
different types of solutions. In addition, special difficulties and common mistakes in
modelling have been discussed.

1.0. Introduction
Groundwater modelling is a way to represent a system in another form to investigate the
response of the system under certain conditions, or to predict the behaviour of the system in
the future.
Groundwater modelling is a powerful tool for water resources management, groundwater
protection and remediation. Decision makers use models to predict the behaviour of a
groundwater system prior to implementation of a project or to implement a remediation

*
E-mail address: Baalousha@web.de, Phone: +64 (0) 6- 8338012, Fax: +64 (0) 6- 8353601
150 Husam Baalousha

scheme. Clearly, it is a simple and cheap solution compared to project establishment in


reality.
By definition, models simplify reality, and are, therefore, imperfect. The famous
statistician George Box insisted, “ all models are wrong, but some are useful” (Box and
Draper 1987). Applicability of any model and its usage depends on the objectives of that
model. Though they are imperfect, models are very useful in hydrogeology. It is a challenge
to the modeller to represent the real word problem in a simplified form without compromising
the accuracy or making invalid assumptions. Modellers try to get the best representation of
reality by collecting as much data as possible and feeding the models with new data.
Groundwater models can be classified into three categories: physical, analogue or
mathematical. Solution of mathematical models can be either analytical or numerical.
Analytical methods do not require much data, but their application is limited to simple
problems. Numerical solutions can handle more complicated problems than analytical
solutions. With the rapid development of computer processors and increasing speed,
numerical modelling has become more effective and easy to use.
The most commonly used numerical modelling approaches are the “finite difference”
method and the “finite element” method. Each method has its advantages and limitations.
Depending on the problem of concern and the objectives of modelling, the appropriate
modelling approach can be selected. Finite difference method can produce different results to
finite element method if the problem of concern is complicated. Modelling approach is not
the only factor that affects the model’s results. Other factors like boundary conditions, initial
condition, time and space discritisation, and quality of data influence the results.
This chapter outlines the stepwise methodology of groundwater modelling, differences
between modelling approaches and difficulties accompany groundwater modelling. Common
mistakes in groundwater modelling are also discussed.

2.0. Modelling Approach


Groundwater Models can be simple, like one-dimensional analytical solutions or spreadsheet
models (Olsthoorn, 1985), or very sophisticated three-dimensional models. It is always
recommended to start with a simple model, as long as the model concept satisfies modelling
objectives, and then the model complexity can be increased (Hill 2006). Regardless of the
complexity of the model being used, the model development is the same.
The stepwise methodology of groundwater modelling is shown in Figure 1. The first step
in modelling is identification of model objectives. Data collection and processing is a key
issue in the modelling process. The most essential and fundamental step in modelling,
however, is model conceptualization. Calibration, verification and sensitivity analysis can be
conducted after model completion and the first run. The following sections explain in detail
each step in groundwater modelling.
Fundamentals of Groundwater Modelling 151

Figure 1. Stepwise methodology of groundwater modelling.


152 Husam Baalousha

2.1. Objectives of Modelling

Groundwater models are normally used to support a management decision regarding


groundwater quantity or quality. Depending on the objectives of modelling, the model extent,
approach and model type may vary.
Groundwater models can be interpretive, predictive or generic. Interpretive models are
used to study a certain case and to analyse groundwater flow or contaminant transport.
Predictive models are used to see the change in groundwater head or solute concentration in
the future. Generic models are used to analyse different scenarios of water resource
management or remediation schemes.
Objectives of groundwater modelling can be listed as:

• Prediction of groundwater flow and groundwater head temporally and spatially.


• Investigating the effect of groundwater abstraction at a well on the flow regime and
predicting the resulting drawdown.
• Investigating the effect of human activities (e.g. wastewater discharge, agricultural
activities, landfills) on groundwater quality.
• Analysis of different management scenarios on groundwater systems, quantitatively
and qualitatively.

Depending on the objectives of study and the intended outcome, selection of model
approach and data requirements can be made to suit the area of study and the objectives. For
example, if the objective is a regional groundwater flow assessment, then a coarse model may
satisfy this objective, but if the area of study is small then a fine-grid model with high data-
density should be used.

3.0. Conceptual Model


A conceptual model is a descriptive representation of a groundwater system that incorporates
an interpretation of the geological and hydrological conditions. Information about water
balance is also included in the conceptual model. It is the most important part of groundwater
modelling and it is the next step in modelling after identification of objectives.
Building a conceptual model requires good information on geology, hydrology, boundary
conditions, and hydraulic parameters. A good conceptual model should describe reality in a
simple way that satisfies modelling objectives and management requirements (Bear and
Verruijt 1987). It should summarise our understanding of water flow or contaminant transport
in the case of groundwater quality modelling. The key issues that the conceptual model
should include are:

• Aquifer geometry and model domain


• Boundary conditions
• Aquifer parameters like hydraulic conductivity, porosity, storativity, etc
• Groundwater recharge
• Sources and sinks identification
Fundamentals of Groundwater Modelling 153

• Water balance

Once the conceptual model is built, the mathematical model can be set-up. The
mathematical model represents the conceptual model and the assumptions made in the form
of mathematical equations that can be solved either analytically or numerically.

3.1. Boundary Value Problem

Mathematical models are all based on the water balance principle. Combining the mass
balance equation and Darcy’s Law produces the governing equation for groundwater flow.
The general equation that governs three-dimensional groundwater steady-flow in
isotropic, homogeneous porous media is:

∂ 2h ∂ 2h ∂ 2h
+ + =0 Equation (1)
∂x 2 ∂y 2 ∂z 2

where h is the groundwater head. This equation is also called the Laplace equation and it has
many applications in physics and hydromechanics. Solving Equation (1) requires knowledge
of boundary conditions to get a unique solution. For this reason, Equation (1) is called a
boundary value problem. So the boundary conditions delineate the area or the domain where
the boundary value problem is valid.

Box 1: Conceptual model: questions to answer


• Is there enough hydrogeological data to describe the geometry of the aquifer/s in
the area of study?
• Should the model be one, two or three-dimensional?
• Is the aquifer/s homogeneous? isotropic?
• What are the sources and sinks?
• What are the sources of contamination (if applicable)?
• Do the boundaries stay the same over time?

3.2. Boundary Conditions

Identification of boundary conditions is the first step in model conceptualisation. Solving of


groundwater flow equations (partial differential equations) requires identification of boundary
conditions to provide a unique solution. Improper identification of boundary conditions
affects the solution and may result in a completely incorrect output. Boundary conditions can
be classified into three main types:

• Specified head (also called Dirichlet or type I boundary). It can be expressed in a


mathematical form as: h (x,y,z,t)=constant
• Specified flow (also called a Neumann or type II boundary). In a mathematical form
it is: ∇h (x,y,z,t)=constant
154 Husam Baalousha

• Head-dependent flow (also called a Cauchy or type III boundary). Its mathematical
form is: ∇h (x,y,z,t)+a*h=constant (where “a” is a constant).

In addition to the above-mentioned types there are other sub-types of boundaries. These
will be explained later.
In groundwater flow problems, boundary conditions are not only mathematical constraint,
they also represent the sources and sinks within the system (Reilly and Harbaugh 2004).
Selection of boundary conditions is critical to the development of an accurate model (Franke
et. al. 1987).
It is preferable to use physical boundaries when possible (e.g., impermeable boundaries,
lakes, rivers) as the model boundaries because they can be readily identified and
conceptualised. Care should be taken when identifying natural boundaries. For example
groundwater divides are hydraulic boundaries and can shift position as conditions change in
the field. If water table contours are used to set boundary conditions in a transient model, in
general it is better to specify flux rather than head. In transient simulation, if transient effects
(e.g. pumping) extend to the boundaries, a specified head acts as an infinite source of water
while a specified flux limits the amount of water available. If the groundwater system is
heavily stressed, boundary conditions may change over time. For this reason, boundary
conditions should be continuously checked during simulation.

3.2.1. Examples of Different Boundaries

Reilly (2001) has surveyed different types of physical features and their equivalent
mathematical representations. Figure 2 shows different types of boundaries. These different
boundaries are briefly described as follows:

Constant head boundary: This is a special case of a specified head boundary, which
occurs where part of the boundary surface of an aquifer coincides with a surface of essentially
constant head (Franke et. al. 1987). Constant head boundaries assume that the head is
constant over time. Lines ABC and EFG in Figure 2 are examples of constant head
boundaries, where part of the aquifer occurs underneath a reservoir.

Specified head boundary: This is a generalised form of constant head boundary. This
occurs when head can be specified as a function of time and location. Rivers and streams,
which are in hydraulic connection with an aquifer, are examples of specified head boundary.

No flow boundary: This is a special case of a specified flux boundary. This occurs at a
line normal to streamline (i.e. normal to flow direction). This case normally occurs where
impermeable media exist. Line HI in Figure 2 represents a no-flow boundary. A water divide
can be used as a no-flow boundary but with caution, as position of the water divide may move
with time as a result of stresses on the aquifer.

Specified flux boundary: This is a generalised case of a no-flow boundary. This occurs
when flow across the boundary can be specified in time and location. An example of a
specified flux boundary is recharge across the water table in a phreatic aquifer. Line CD in
Figure 2 is a specified flux boundary.
Fundamentals of Groundwater Modelling 155

Head-dependent flux boundary: This occurs when flux across a boundary depends on
head adjacent to that boundary. A semi-confined aquifer, where the water head depends on
the flux through the semi-confining layer, is an example of this type of boundary. This can be
represented by lines ABC and EFG in Figure 2.

Free-surface boundary: The water table and the fresh-saline water interface in a coastal
aquifer are examples of free-surface boundaries. Line CD in Figure 2 represents a free-surface
boundary. Pressure head at free-surface boundary is always zero and the total head equals
elevation head.

Seepage face boundary: This occurs at the boundary between saturated flow and the
atmosphere. The face of a landfill dam, as shown by line DE in Figure 2 is an example of a
seepage face boundary.

Figure 2. Different types of boundaries.

Box 2: Boundary conditions comments


• Always use natural boundaries when possible.
• Boundary conditions always influence a steady state solution but may not
influence a transient solution.
• A steady state solution with all specified flux boundary conditions (including no
flow) without specified head or head-dependent internal boundaries may not
converge or may not give a unique solution.
• A specified head boundary acts as an infinite source or sink.
• A water divide should be used as a no-flow boundary with caution.

4.0. Types of Models


There are different types of models to simulate groundwater movement and contaminant
transport. In general, models can be classified into three categories: physical, analogue and
156 Husam Baalousha

mathematical models. The latter type can be classified further depending on the type of
solution.

4.1. Physical Models

Physical models (e.g. sand tanks) depend on building models in the laboratory to study
specific problems of groundwater flow or contaminant transport. These models can
demonstrate different hydrogeological phenomena like the cone of depression or artesian
flow. In addition to flow, contaminant movement can be investigated through physical
models. Though they are useful and easy to set-up, physical models cannot handle
complicated real problems.

4.2. Analogue Models

The equation which describe groundwater flow in isotropic homogenous porous media is
called the Laplace’s Equation (Equation (1)). This equation is very common in many
applications in physical mathematics like heat flow, and electricity. Therefore, comparison
between groundwater flow and other fields where the Laplace equation is valid, is possible.
The most famous analogue model is the flow of electricity. The electric analogue is based
on the similarity between Ohm’s law of electric current flow and Darcy’s law of groundwater
movement. As electric current moves from high voltage to lower voltage, so does the
groundwater, which moves from high head to lower head.
Simple analogue models can easily be setup to study the movement of groundwater flow.
More detailed information on analogue models is available (Verruijt, 1970, Anderson and
Woessner, 1992, Strack 1989; Fetter 2001).

4.3. Mathematical Models

Mathematical models are based on the conceptualisation of the groundwater system into a set
of equations. These equations are formulated based on boundary conditions, initial conditions,
and physical properties of the aquifer. Mathematical models allow an easy and rapid
manipulation of complex models.
Once the mathematical model is set-up, the resulting equations can be solved either
analytically, if the model is simple, or numerically.

5.0. Types of Model Solutions


As discussed in the preceding sections, the mathematical models can be solved either
analytically or numerically. Some approaches use a mixture of analytical and numerical
solutions. The following sections briefly discuss the main types of solutions used in
groundwater modelling.
Fundamentals of Groundwater Modelling 157

5.1. Analytical Solutions

Analytical solutions are available only for simplified groundwater and contaminant transport
problems. They were developed before the use of numerical models. The advantages of
analytical solutions are that they are easy to apply and produce continuous and accurate
results for simple problems. Unlike numerical solutions, analytical solutions give a
continuous output at any point in the problem domain (Figure 3). However, analytical
solutions make many assumptions like isotropy and homogeneity of an aquifer, which are not
valid in general. Analytical solutions; therefore, cannot deal with complex groundwater
systems. Examples of analytical solutions are the Toth solution (Toth, 1962) and Theis
equation (1941). More details on analytical solutions of groundwater problems can be found
in Bear (1979) and Walton (1989).

5.2. Numerical Solutions

Because analytical solutions of partial differential equations (PDE) implies many


assumptions, simplifications and estimations that do not exist in reality, they cannot handle
complicated real problems. Numerical methods were developed to cope with the complexity
of groundwater systems.
Numerical models involve numerical solutions of a set of algebraic equations at discrete
head values at selected nodal points (Figure 3). The most widely used numerical methods are
finite difference and finite element methods. Other methods have been developed, such as the
boundary-element method.

60
Numerical
50
Groundwater head (h)

Analytical

40

30

20

10

0
0 100 200 300 400 500 600 700 800 900 1000
Distance (x)

Figure 3. Analytical versus numerical solutions for a 1-D groundwater flow problem.

5.2.1. Finite Difference Method

Finite difference method (FDM) has been widely used in groundwater studies since the early
1960s. FDM was studied by Newton, Gauss, Bessel and Laplace (Pinder and Gray 1977).
158 Husam Baalousha

This method was first applied in petroleum engineering and then in other fields. The finite
difference method depends on the estimation of a function derivative by a finite difference
(Figure 4). The finite difference approximation is given by:

dy Δy Δy
= lim ≈ Equation (2)
dx Δx →∞ Δx Δx

The accuracy of the method depends on grid size and uniformity. The approximation of
the derivative improves as grid spacing approaches zero; however, numerical dispersion and
truncation error increases. There are three different methods of finite difference
approximation: forward, backward and central difference, depending on the way the finite
difference is implemented. The central difference gives the best results as the truncation error
is of second order O (Δx)2 (Pinder and Gray, 1970).

Figure 4. Finite difference approximation.

The general governing equation for transient, heterogeneous, and anisotropic conditions
is given by:

∂ ∂h ∂ ∂h ∂ ∂h ∂h
( K x ) + ( K y ) + ( K z ) = Ss −W Equation (3)
∂x ∂x ∂y ∂y ∂z ∂z ∂t

where Kx, Ky, and Kz are hydraulic conductivity in x, y and z directions, respectively. W is
the sink or source term and Ss is specific storage.
Fundamentals of Groundwater Modelling 159

For simplicity, consider a one-dimensional case of Equation (3) and solve for h using the
finite difference method. This yields:

∂ 2 h ∂ ⎛ ∂h ⎞ 1 ⎡ hi +1 − hi hi − hi −1 ⎤ hi −1 − 2hi + hi +1
= ⎜ ⎟= − =
∂x 2 ∂x ⎝ ∂x ⎠ Δx ⎢⎣ Δx Δx ⎥⎦
Equation (4)
(Δx) 2

where hi, hi+1 are the head at node i, and node i+1 respectively (Figure 5). Irregular spacing
can be used to increase accuracy at selected areas of the grid, but this increases associated
error more than regular-spaced grids. As a rule of thumb for expanding a finite difference
grid, the maximum multiplication factor should not be higher than 1.5

Figure 5. The finite difference method.

Figure 6. Discritization of model domain into a finite difference grid.


160 Husam Baalousha

The advantages of the finite difference method are that it is easy to implement, well
documented and produces reasonably good results. However, finite difference method has
some disadvantages. The main disadvantage is that it does not fit properly to an irregular
model boundary (Figure 6). In addition, the distribution of grids, their size, and whether they
are of equal size highly affects the accuracy and computation effort. Output accuracy of the
finite difference method is not good in the case of solute transport modelling. Mass balance is
not guaranteed if conductivity or grid spacing varies (Cirpka 1999).
The most widely used finite difference based groundwater model is MODFLOW
(Harbaugh and McDonald 1996).

Box 3: Considerations in selecting the size of the nodal spacing in grid or mesh
design
• Variability of aquifer characteristics (e.g.. conductivity, storativity).
• Variability of hydraulic parameters (e.g. recharge, pumping).
• Curvature of the water table.
• Desired detail around sources and sinks (e.g., rivers).
• Vertical change in head (vertical grid resolution/layers).

5.2.2. Finite Element Method

The basis of the finite element method is solving integral equations over the model domain.
When finite element method is substituted in the partial differential equations, a residual error
occurs. The finite element method forces this residual to go to zero.
There are different approaches for the finite element method. These are: basis functions,
variational principle, Galerkin’s method, and weighted residuals. Detailed description of each
method can be found in Pinder and Gray (1970).
Finite element method discritises the model domain into elements (Figure 7). These
elements can be triangular, rectangular, or prismatic blocks. Mesh design is very important in
the finite element method as it significantly affects the convergence and accuracy of the
solution. Mesh design in the finite element method is an art more than a science, but there are
general rules for better mesh configuration. It is highly recommended to assign nodes at
important points like a source or sink, and to refine mesh at areas of interest where variables
change rapidly. It is better to keep the mesh configuration as simple as possible. In the case of
triangular mesh, a circle intersecting vertices should have its centre in the interior of the
triangle.
The weighted residual method is being widely used in groundwater finite element
problems. The Russian engineer B. G. Galerkin introduced this method in 1915 (Pinder and
Gray 1970). To illustrate the weighted residual approach, consider a groundwater or solute
transport problem. This problem over a domain B can be written as:

L(φ ( x, y, z )) − F ( x, y, z ) = 0 Equation (5)

Where L is a differential operator, φ(x,y,z) is the dependent variable (i.e. groundwater


head) and F(x,y,z) is a known function.
Fundamentals of Groundwater Modelling 161

The weighted residual method replaces the dependent variable φ(x,y,z) by an


approximation function φˆ( x, y, z ) . The later approximation function is made up of a linear
combination of a new function that satisfies the boundary conditions of the main problem. It
can be written as:

m
φˆ( x, y, z ) = ∑ N i ( x, y, z ) • φi Equation (6)
i =1

where Ni is an interpolation function, φi is the unknown nodal value of dependent variable at


node i, and m is the number of nodes.
Because φˆ( x, y, z ) is an approximation, there will be a residual R(x,y,z) at each node.
This residual is given by:

R ( x, y, z ) = L((φˆ( x, y, z ) − F ( x, y, z ) ≠ 0 Equation (7)

The weighted residual method forces the residual in Equation (7) to go to zero. This
requires:

∫∫∫W ( x, y, z ) •R( x, y, z )dxdydz = 0


B
Equation (8)

Where W (x,y,z) is a weighting function and B is the problem domain. Equation (8) can
be written in terms of approximation function as follows:

∫∫∫W ( x, y, z)[ L(φˆ( x, y, z )) − F ( x, y, z )]dxdydz = 0


B
Equation (9)

In case of a steady state, two-dimensional groundwater flow problem, Equation (9) can
be written as:

⎡ ∂ ⎛ ∂hˆ ⎞ ∂ ⎛ ∂hˆ ⎞⎤
∫∫B W ( x , y , z ) ⎢ ⎜⎜ k x ⎟ + ⎜ky
⎟ ⎜
⎟⎥ dxdy = 0
⎟ Equation (10)
⎢⎣ ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎠⎥⎦

To solve Equation (10), the weighting function W (x,y,z) needs to be identified. There are
different methods of weighting residuals in addition to Galerkin’s approach. More details on
weighting residuals methods can be found in Gray and Pinder (1970) and Reddy (2006).
The main characteristics of the finite element method are: properties and source/sink are
assigned at nodes, nodes are located at flux boundaries, and it suites aquifer anisotropy better
than FDM. Advantages of this method include: a better mesh configuration, which suites
irregular model boundaries, anisotropy is well incorporated, the governing system of
equations is symmetric and irregular shapes can be used to represent elements.
162 Husam Baalousha

Figure 7. Discritization of the model domain into a finite element mesh.

The finite element method has some disadvantages. The finite element mesh is not easy
to build and consumes time, especially in complicated problems. Also, there is less
documentation on the finite element method compared to finite difference method. Unlike the
finite difference method, mass balance in the finite element method can be achieved for the
entire domain but not for every element.
The most well known finite element based groundwater models are Feflow (Wasy, 2005),
Femwater (Lin, et. al. 1997), and MODFE (Torak 1993).

Box 4: Finite element or finite difference?


Finite element Finite difference

Model boundary Incorporates irregular and Difficult to incorporate


curved boundaries irregular boundaries

Nodes At vertices and flux boundary Nodes located at centre of


cell
Mesh/grid building Difficult to generate an Easy to build a finite
efficient mesh difference grid

Anisotropy Easily incorporated Difficult to incorporate

Accuracy Acceptable accuracy More accurate especially in


solute transport modelling

Computation time Good and acceptable time Computation time can be


requirement long in 3D problems.
Fundamentals of Groundwater Modelling 163

6.0. Model Calibration


After the first run of a model, model results may differ from field measurements. This is
expected because modelling is just a simplification of reality and approximations and
computational errors are inevitable.
The process of model calibration is aimed at fine-tuning the model results to match the
measurements in the field. In a groundwater flow models, the resulting groundwater head is
forced to match the head at measured points. This process requires changing model
parameters (i.e. hydraulic conductivity or groundwater recharge) to achieve the best match.
The calibration process is important to make the model predictive and it can also be used for
inverse modelling.
To illustrate the calibration process of a groundwater flow model, consider the
groundwater head measurements (hob)i at the observation point i. The simulated head at the
same point is (hsim)i. The root mean square error of the residual is given by:

1/ 2
⎡1 n ⎤
RMSE = ⎢ ∑ (hob − hsim ) 2 ⎥ Equation (11)
⎣ n i =1 ⎦

Calibration involves an optimization process to minimize the RMSE given in Equation


(11). To get a well-calibrated model, proper site characterisation and ample data are required.
Otherwise, the calibrated model will only be valid for a set of conditions and not for any
condition. Calibration can be manually done or automatically. Software like PEST (Doherty
et. al. 1994) and UCODE (Poeter and Hill 1994) can be used for automatic calibration.

Box 4: The calibrated model should satisfy:


• A good match between measured and modelled head.
• A good water mass balance.
• The groundwater gradient from the model is similar to the observed gradient in
the
field.
• Similar behaviour for any dataset.

7.0. Model Verification and Validation


The term “validation” is not completely true when used in groundwater modelling. Oreskes
et. al. (1994) asserted it is impossible to validate a numerical model because modelling is only
approximation of reality.
Model verification and validation is the next step after calibration. The objective of
model validation is to check if the calibrated model works well on any dataset. Because the
calibration process involves changing different parameters (i. e. hydraulic conductivity,
recharge, pumping rate etc.) different sets of values for these parameters may produce the
same solution. Reilly and Harbaugh (2004) concluded that good calibration did not lead to
good prediction. The validation process determines if the resulting model is applicable for any
164 Husam Baalousha

dataset. Modellers usually split the available measurement data into two groups; one for
calibration and the other for validation.

8.0. Sensitivity Analysis


Sensitivity analysis is important for calibration, optimisation, risk assessment and data
collection. In regional groundwater models, there are a large number of uncertain parameter.
Coping with these uncertainties is time-consuming and requires considerable effort.
Sensitivity analysis indicates which parameter or parameters have greater influence on the
output.
Parameters with high influence on model output should get the most attention in the
calibration process and data collection. In addition, the design of sampling location, and
sensitivity analysis can be used to solve optimisation problems.
The most common method of sensitivity analysis is the use of finite difference
approximations to estimate the rate of change in model output as a result of change in a
certain parameter. The Parameter Estimation Package “PEST” uses this method (Doherty et.
al. 1994).
Some other more efficient methods of sensitivity analysis have been used. Automatic
differentiation has been used for sensitivity analysis in groundwater models and it produces
precise output compared to finite difference approximations (Baalousha 2007).

9.0. Uncertainty Analysis


Uncertainty in groundwater modelling is inevitable for a number of reasons. One source of
uncertainty is the aquifer heterogeneity. Field data has uncertainty. Mathematical modelling
implies many assumptions and estimations, which increase the uncertainty of the model
output (Baalousha and Köngeter 2006).
There are different approaches to incorporate uncertainty in groundwater modelling. The
most famous approach is stochastic modelling using the Monte Carlo or Quasi Monte Carlo
method (Kunstmanna and Kastensb. 2006: Liou, T. and Der Yeh, H. 1997). The problem with
stochastic models is that they require a lot of computations, and thus they are time
consuming. Some modifications have been done on stochastic models to make them more
deterministic, which reduces computational and time requirements. Latin Hypercube
Sampling is a modified form of Monte Carlo Simulation, which considerably reduces the time
requirements (Zhang and Pinder 2003).

10.0. Common Mistakes in Modelling


A major mistake in modelling is conceptualisation. If the conceptual model is incorrect, the
model output will be incorrect regardless of data accuracy and modelling approach. A good
mathematical model will not resurrect an incorrect conceptual model (Zheng and Bennet,
2002).
Fundamentals of Groundwater Modelling 165

In all models, it is necessary to identify a certain reference elevation for all head so that
the model algorithm can converge to a unique solution (Franke et. al. 1987).
Boundary conditions should be treated with care, especially in a steady state simulation.
Sometimes boundary conditions change during simulation and become invalid. A model with
hydraulic boundary conditions will be invalid if stresses inside or outside the model domain
cause the hydraulic boundaries to shift or change. Therefore, boundary conditions should be
monitored at all times to ensure they are valid.
Model parameterisation is a common mistake in modelling. Theoretical values of
hydraulic properties or groundwater recharge should never substitute field data and field
investigation. Assumptions like isotropy and homogeneity should not be used without support
from field investigation.
Selection of the model code is important to obtain a good solution. Different codes
involve different mathematical settings that suit a certain problem. The selected code should
consider characteristics of the area of interest and the objectives of modelling.
Models can be well calibrated and match well with the measured values, but have an
incorrect mass balance. This can be a result of an improper conceptual model.

Reference
Anderson, M. and Woessner, W. (1992) Applied groundwater modeling. Elsevier. 381p.
Baalousha, H. (2007) Application of Automatic Differentiation in Groundwater Sensitivity
Analysis. In Oxley, L. and Kulasiri, D. (eds) MODSIM 2007 International Congress on
Modelling and Simulation. Modelling and Simulation Society of Australia and New
Zealand, December 2007, pp. 2728-2733. ISBN : 978-0-9758400-4-7.
Baalousha, H and Köngeter, J. (2006) Stochastic modelling and risk analysis of groundwater
pollution using FORM coupled with automatic differentiation. Advances in Water
Resources,. 29(12): 1815-1832
Bear, J. (1979) Hydraulics of Groundwater. McGraw-Hill, New York.. 567p.
Bear, J. and Verruijt, A. (1987) Modeling Groundwater Flow and Pollution. Springer, 432p.
Box, G. and Draper, N. (1987) Empirical Model-Building and Response Surfaces, 669p.,
Wiley.
Cirpka, O. 1999 Numerical methods of groundwater flow and transport. Technical report.
Stanford University, Department of Civil and Environmental Engineering.
Doherty, J., Brebber, L. and Whyte, P. (1994) PEST - Model-independent parameter
estimation. User’s manual. Watermark Computing. Australia
Fetter, C.W. (2001) Applied Hydrogeology. Prentice Hall. 4th ed.
Franke, O.L., Reilly, T.E. and Bennett, G.D., (1987) Definition of boundary and initial
conditions in the analysis of saturated ground-water flow systems – An introduction:
Techniques of Water-Resources Investigations of the United States Geological Survey,
Book 3, Chapter B5, 15 p
Harbaugh, A. and McDonald, M. (1996) User's documentation for MODFLOW-96, an update
to the U.S. Geological Survey modular finite-difference ground-water flow model: U.S.
Geological Survey Open-File Report 96-485, 56 p.
Hill, Mary. (2006) The practical use of simplicity in developing groundwater models. Ground
water Journal, 44(6): 775-781.
166 Husam Baalousha

Kunstmanna, H. and Kastensb, M. (2006) Direct propagation of probability density functions


in hydrological equations. Journal of Hydrology , 325(1-4): 82-95
Lin, Hsin-Chi J. , Richards, David R. ; Yeh, Gour-Tsyh , Cheng, Jing-Ru and Cheng, Hwai-
Ping (1997) FEMWATER: A Three-Dimensional Finite Element Computer Model for
Simulating Density-Dependent Flow and Transport in Variably Saturated Media. Army
Engineer Waterways experiment station vicksburg ms coastal hydraulics lab.
Liou, T. and Der Yeh, H. (1997) Conditional expectation for evaluation of risk groundwater
flow and solute transport: one-dimensional analysis. Journal of Hydrology, 199(3-4):
378-402
Olsthoorn, T. (1985) the power of the electronic worksheet- modelling without special
programs. Ground Water Journal, 23: 381-390
Oreskes, N., Shrader-Frechette, K. and Belitz, K. (1994) Verification, Validation, and
Confirmation of Numerical Models in the Earth Sciences. Science, 263(5147): 641-646.
Pinder, G. and Gray, W. (1970) Finite element simulation in surface and subsurface
hydrology. Academic Press Inc. 295p.
Poeter, EP. and Hill, MC. (1998) Documentation of UCODE, a computer code for universal
inverse modeling, U.S. Geological Survey, Water-Resources Investigations Report 98-
4080
Reddy, J. (2006) An Introduction to the finite element method. McGraw-Hill.912p.
Reilly, T. (2001) System and Boundary conceptualization in ground-water flow simulation.
Techniques of water resources investigations of the U.S. Geological Survey. Book 3,
Applications of Hydraulics. Chapter B8. Department of Interior,. U.S. Geological Survey.
Reilly, T. and Harbaugh, A. (2004) Guidelines for evaluating Ground-Water flow. Scientific
Investigations Report 2004-5038. U.S. Department of Interior,. U.S. Geological Survey.
Strack, ODL. (1989) Groundwater Mechanics. National Water Well Association, Dublin,
Ohio. 732p
Theis, CV. (1941) The effect of a well on the flow of a nearby stream. American Geophysical
Union Transactions 22 (3): 734-738
Torak, L.J. (1993) A MODular Finite-Element model (MODFE) for areal and axisymmetric
ground-water-flow problems, part 1--model description and user's manual: U.S.
Geological Survey Techniques of Water-Resources Investigations, book 6, chap. A3.
Toth, J. (1962) A theory of groundwater motion in small drainage basins in central Alberta:
Journal of Geophysical Research, 67(11): 4375-4387.
Verruijt, A. (1970) Theory of groundwater flow. Macmillan and Co. LTD 190p.
Walton, W. (1989) Analytical Ground Water Modeling. Lewis Publishers, Chelsea, Michigan.
Wasy GmbH. (2005) Feflow: finite element subsurface flow and transport simulation system.
Reference Manual. Wasy GmbH, Berlin.
Zhang, Y. and Pinder, G. (2003) Latin Hypercube lattice sampling selection strategy for
correlated random hydraulic conductivity fields. Water Resources Research 39(8) doi:11-
1/11-3.
Zheng, C., and Bennett, G. (2002) Applied Contaminant Transport Modeling. Wiley
InterScience: New York, NY. 2nd ed. 621 p.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 167-186 © 2008 Nova Science Publishers, Inc.

Chapter 6

CONTAMINANTS IN GROUNDWATER
AND THE SUBSURFACE

Z. Yu1,*, Y. Huang1, A. Baron1, X. Chen1, C. Yang1,


D. Kreamer1 and M. Johnson2
1
Department of Geoscience, University of Nevada Las Vegas
Las Vegas, NV 89154
2
Virgin Valley Water District, 500 Riverside Road, Mesquite, NV 89027

Abstract
Remediation of groundwater contamination is becoming increasingly important as
groundwater resources are more extensively relied upon for industrial and drinking water
supplies. Subsurface contaminants can be divided into three general categories: 1) microbial
pathogens, which can be filtered out by some subsurface media and therefore pose the greatest
risk in aquifers with large cracks or conduits, 2) inorganic compounds such as major anions
and cations, metals, and radioactive compounds, whose mobility in groundwater depends on
their interaction with the geologic material, and 3) organic chemicals, including many
pesticides and nonaqueous phase liquids such as petroleum products and solvents. The
movement of contaminants through the subsurface is governed by hydrological, physical,
chemical, biological and transport processes, all of which are explained in detail. Investigation
of groundwater contamination should begin with noninvasive techniques, including review of
historical records, maps and photographs, and some geophysical methods. Contaminant
plumes can then be delineated using soil samples and monitoring wells. Numerical models, as
long as they are developed via a rigorous procedure based on specific questions and available
data, can integrate spatial and temporal variables to more accurately describe the complexity
of subsurface contaminant fate and transport. Various unsaturated and saturated zone models
are described. Remediation in its most basic forms consists of source containment, excavation
of soil in the unsaturated zone, and pump-and-treat methods for water in the saturated zone.
These methods are often impractical, which has led to the ongoing development of
supplemental and alternative methods, described briefly herein.

*
E-mail address: zhongbo.yu@unlv.edu. Phone: 702-895-2447.Fax: 702-895-4064. Contact author: Zhongbo Yu,
Department of Geoscience, University of Nevada Las Vegas, 4505 Maryland Parkway, Las Vegas, NV 89154-4010
168 Z. Yu, Y. Huang, A. Baron et al.

Keywords: Contamination; fluid flow; contaminant transport; chemical reactions;


contamination detection and monitoring; modeling; remediation.

1. Introduction
Although water is indispensable to life, we have often failed to recognize it as a precious
resource. Groundwater resources, increasingly used as primary sources for industrial and
drinking water supplies, have been threatened over the past century not only by excessive
withdrawal but also by contamination resulting from past and present industrial, agricultural,
and commercial activities. This contamination has consisted of the deliberate use of
chemicals in industry and agriculture, waste disposal, unintended transport accidents, leaking
storage facilities, and other human activities. There are estimated to be 300,000 contaminated
soil and groundwater sites requiring expenditure of hundreds of billions of dollars for
remediation and monitoring in the United States alone (EPA, 1991). Equally serious problems
of soil and groundwater contamination occur in many other industrialized nations and,
increasingly, in developing countries.
There is a growing concern about the impacts of soil and groundwater contamination on
human health and hydrological, biological, and ecological systems. Environmental laws and
public outcry have galvanized efforts to clean up contaminated groundwater as the number of
serious incidents of human illness and other damage to the biosphere has increased. In the
United States, the Compensation and Liability Act of 1980 (referred to as CERCLA or the
Superfund Act), the Resource Conservation and Recovery Act (RCRA), the Safe Drinking
Water Act, and other federal legislation and State laws have established regulations,
guidelines, and financial responsibility for groundwater restoration.
In the past, groundwater contamination was addressed primarily by monitoring water
quality and treating extracted water to remove or reduce contaminants (called
pump-and-treat). Contaminant levels in groundwater typically decrease very slowly and, over
the past two decades, environmental agencies and scientists have become more aware of the
costs of supplementing pump-and-treat groundwater cleanup . The traditional pump-and-treat
approach has also become less acceptable in an era of rises in water demand, contaminant
sources, and awareness of health and ecological risks. Regulatory agencies now endorse an
integrated approach to dealing with soil and groundwater contamination that emphasizes
contaminant mass removal and source control. However, this integrated approach requires
more research to better understand the fate and transport of contaminants in the soil and
groundwater.
This article begins with an introduction of the types of contaminants in the environment
and some fundamental hydrologic processes that control contaminant storage and transport.
This knowledge is prerequisite for the development of strategies for contamination prevention
and remediation and for the construction of conceptual mathematical models used in research
and as regulatory and management tools. Various technologies for detecting and monitoring
contaminant movement are described. The authors then discuss the way in which numerical
models are developed and used to study contaminant transport in the soil and groundwater.
The article concludes with a discussion of some traditional methods and new technologies for
contaminant remediation.
Contaminants in Groundwater and the Subsurface 169

2. Subsurface Contaminants
There are many types of pollutants that find their way into the soil, the remainder of the
unsaturated zone, and groundwater. Their sources can vary greatly. For example, microbial
contamination can emanate from septic tank leach fields, or seepage from disposed municipal
sewage treatment effluent. Water percolating through landfills can create leachate, which
adds chemical contaminants to soil and groundwater. Industrial processes can discharge
inorganic and organic chemicals through spillage, improper storage procedures, or direct
release. Major classes of contaminants from these sources include: 1. microbial pathogens
(disease-causing agents), 2. inorganic compounds such as major anions and cations, metals,
and radioactive compounds, and 3. organic chemicals including many pesticides and
nonaqueous phase liquids such as petroleum products and solvents.
Worldwide, disease contracted from microbial pathogens is one of the leading causes of
mortality. In well water, microbial pathogens pose the greatest threat in situations where
groundwater flows through large cracks or conduits in the subsurface and very little particle
filtering occurs. Pathogens can migrate quickly and move great distances in these types of
rocks. Aquifer types that contain large conduits include hard rocks where groundwater flow
occurs in fractures, and limestones (karst systems), where flow is primarily through
dissolution cavities. Fortunately, there are many groundwater systems in which filtering does
occur, including unconsolidated sediments and most sandstones. This filtration prevents
migration of all but the smallest particles. Pathogenic organisms come in different shapes and
sizes, with the largest being the eucaryotes, complicated microorganisms that are often
motile. Eucaryotic cells, including most protozoans, have cellular volumes in the range of
about 20 - 50,000 um3, and are quickly filtered and stopped in many subsurface
environments. Procaryotic cells, including bacterial pathogens, are smaller, with an
approximate range of 0.01 - 5 um3 in cellular volume, but are still inhibited in subsurface
movement. Viruses, which are simply nucleic acids surrounded by a protein coat, are the
smallest pathogens (0.00001 - 0.01 um3) and therefore the most mobile in the subsurface.
Inorganic chemicals constitute another important class of contaminants. They dissolve
into water in varying degrees and often undergo ionization. The surfaces of many minerals
are negatively charged (e.g. clays), repulsing negatively charged anions dissolved in water
and attracting positively charged cations. Anions are more mobile underground than cations
because of this property. The anion nitrate (NO3-), for example, a common by-product of the
degradation of fertilizer, sewage, or animal waste, is quite mobile in groundwater. Nitrate can
be reduced to nitrite (NO2- ) in the gastrointestinal tract of infants when ingested, which in
turn interferes with hemoglobin's ability to transfer oxygen in the baby's bloodstream, causing
methemoglobinemea or "blue baby" syndrome.
There are many important inorganic contaminants that can dissolve in water, including
many ionic species, metals, and radioactive contaminants, but geologic materials can inhibit
their movement. Strong acids and bases react readily with soils and rocks, which reduce their
potential harmfulness and migration. Metals can be contributed to soils and groundwater by
chemical weathering of rocks, mining activities, or the release of industrial by-products.
However, many free metals have low mobility due to sorption on soils, deeper sediments and
rocks. Exceptions include chromium and molybdenum, which can exist as anions and migrate
more readily. These metals change electrical charge (valence) under varying subsurface pH
170 Z. Yu, Y. Huang, A. Baron et al.

and oxygen conditions, forming different chemical species with dramatically different
mobilities. Metals can also mobilize when mixed with chemical complexing or chelating
agents that essentially surround the metal and don't allow it to adsorb to geologic material.
The ability of these species and chemical complexes to move in groundwater is a potential
health problem because some metals have high toxicity (e.g. cadmium, mercury, lead).
Likewise, some inorganic isotopes are radioactive, emitting particles as the isotope decays
and causing cellular damage to an animal that ingests water with radioactive constituents.
There are thousands of organic chemicals that have created groundwater and soil
pollution problems. Although many organic compounds in water are derived from the natural
decomposition of plant and animal material, there are increasing challenges associated with
synthetic organic compounds in groundwater, which have a variety of health related impacts,
such as toxicity, endocrine disruption, carcinogenicity and teratogenicity.
Some of these manufactured organic compounds or synthetic by-products are: pesticides,
refined petroleum products, polychlorinated biphenyls (PCBs), solvents, and trihalomethanes.
Pesticides have a broad category which is made up of insecticides, rodenticides, herbicides,
fungicides and other miscellaneous fumigants, growth depressants and repellents. They
include such chemical groups as the environmentally long-lived chlorinated hydrocarbons
(e.g., DDT, Dieldrin, Chlordane), the more acutely toxic organic phosphates (e.g., Parathion,
Diazinon, Malathion), the systemic carbamates, phenoxyalkanoic acid derivatives (possibly
containing dioxins), substituted ureas, and triazines.
Petroleum products are another broad class of organic compounds, considered light
nonaqueous phase liquids (LNAPLs) because they are less dense than water and less
immiscible in it. Although gasoline typically contains several hundred compounds, the light
aromatic compounds of benzene, toluene, ethyl benzene, and the xylenes are its most soluble
components in water, and therefore the most mobile. Another characteristic of LNAPLs is
that, as they leak into the subsurface, they tend to float on the water table and pool up as a
separate, slowly dissolving phase. The universal storage of fuels in metal underground tanks
with a propensity for corrosion has created LNAPL-related groundwater contamination
problems worldwide.
Dense nonaqueous phase liquids (DNAPLs), another class of immiscible liquids, are
denser than water and therefore generally sink below water tables, making monitoring and
remediation efforts extremely difficult. DNAPLs include the chlorinated solvents that are
generally degreasing agents, PCBs, which were used in electrical transformers and other
applications, and creosote, which is a wood preservative. Even the process of disinfection of
water by chlorination can create carcinogenic trihalomethanes, another organic compound.
Understanding the transport of these contaminants is essential to proper monitoring,
assessment and remediation of the subsurface environment.

3. Hydrologic Processes in the Soil and Groundwater


Water flowing through soil and in groundwater is a part of the larger hydrologic cycle that
encompasses the ocean, atmosphere, and land areas. As we are all aware, drinking water
obtained from groundwater can be easily contaminated by the leakage of contaminants into
the soil from the ground surface. Understanding the various mechanisms of water flow and
Contaminants in Groundwater and the Subsurface 171

contaminant transport along pathways in the soil and groundwater is crucial to contamination
prevention and remediation.
Movement of water through soil as a steady flow can be described with Darcy's equation:

Q ∂h
= q = - K( θ ) (1)
A ∂z

where Q is the volumetric flow rate (cm3/day), A is the cross-sectional area (cm2), q is the
volumetric flow rate per unit surface area (cm/day), K(θ) is the hydraulic conductivity
(cm/day), θ is the volumetric water content (cm3/cm3), h is the hydraulic head, and z is
distance (cm). The specific discharge q in the saturated zone or groundwater is a superficial or
apparent velocity because water only moves through the pore openings making up the surface
area (Domenico and Schwartz, 1990). The more realistic velocity (v) is a volumetric flow rate
per unit area of connected pore space: v = q / ne, where ne is the effective porosity.
Thousands of measurements of hydraulic conductivity have been obtained in the field and
laboratory over the years. Values of hydraulic conductivity can range over 11 orders of
magnitude from 1x10-8 cm/day in unfractured igneous and metamorphic rocks to tens or
thousands of meter per day in gravels and some karstic or reef limstones (Davis, 1969).
Various direct and indirect field and laboratory methods are available for measuring hydraulic
conductivity. In unsaturated soils, the hydraulic conductivity is a function of saturated
hydraulic conductivity and soil water content. The soil water content can be measured by
simply drying samples, by neutron probe, or by time domain reflectometry (TDR).
Various schemes are available for parameterizing soil hydraulic properties (Clapp and
Hornberger, 1978; van Genuchten, 1980; Cosby et al., 1984), and have been widely used in
different soil and hydrological applications. They conceptualize hydraulic conductivity and
soil-water metric potential as functions of the normalized volumetric water content (Se),
which is defined as:

θ -θ r
Se= (2)
θ s -θ r

where θs is the soil water content at saturation, and θr is the residual soil water content. A
simple equation by Clapp and Hornberger (1978) and Cosby et al. (1984) can be used for
relating hydraulic conductivity to soil water content:

K( θ ) = K s ( S e )3+1/λ (3)

where Ks is the saturated hydraulic conductivity (cm/day), and λ is a pore size parameter. A
more versatile equation by Rawls and Brakensiek (1985), relating the hydraulic conductivity
to soil water content, is the following:

2
K( θ ) = K s ( S e )1/2 [1 - (1 - S e1/m )m ] (4)
172 Z. Yu, Y. Huang, A. Baron et al.

where m = λ / (λ+1). Choice of the numerical values of these parameters is critical to


correctly calculating soil water content.
The following equation can be used to describe transient flow in the vertical direction:

∂θ ∂q
=- (5)
∂t ∂z

Combining Equation 5 with Darcy's equation (Equation 1) yields the following, which is
Richards' equation for describing the rate of change of volumetric water content
(Swartzendruber, 1969):

∂θ ∂q ∂ ∂ψ ∂θ ∂K ∂θ
= - = [K ]+ (6)
∂t ∂z ∂z ∂θ ∂z ∂θ ∂z

where ψ is hydraulic potential. Analytic solutions are available for Equation 6 for some well-
defined conditions. In general, however, Richards' equation is solved numerically for the
solution of soil moisture flow flux.
The transformation of precipitation into surface runoff is controlled by the independent
interaction of many spatially variable processes. Several rainfall-runoff generating processes
have been recognized over the years (Dunne, 1978; Freeze, 1980; Beven, 1989). Horton
runoff (Horton, 1933) and Dunne runoff (Dunne and Black, 1970) are perhaps the two most
important conceptual models for the production of surface runoff. Horton runoff is considered
the excess of precipitation intensity over soil infiltration rate at a point (Freeze, 1974):

f(t) = f c + ( f 0 - f c ) e-kt (7)

where f(t) is the infiltration at time t (cm/hr), f0 is the initial infiltration rate (cm/hr), fc is the
constant infiltration rate (cm/hr), and k is a decay constant. Philip (1957) solved Richards’
equation under less restrictive conditions by relating conductivity and diffusivity to the soil
moisture content. Green and Ampt (1911) developed approximate solutions of Richards’
equation for infiltration calculation.
Dunne runoff occurs when precipitation falls over a saturated area or impermeable
surface. This flow is partially responsible for the rapid streamflow response associated with
the expansion of saturated areas (referred as to variable source areas) along streams or valleys
during a storm event (Dunne and Black, 1970). Various models, including TOPMODEL by
Beven and Kirkby (1979), were developed to study variable source areas (Troendle, 1985).

4. Flow and Transport Processes


The fate and transport of contaminants in the surface and subsurface is determined by the
combination of various interactive processes. Major processes controlling the fate and
transport of contaminants in the soil and groundwater system are described in Figure 1.
Contaminants in Groundwater and the Subsurface 173

Although they interact with each other and show a continuous change over time, they can be
classified into three groups:

• Physical and chemical reactions processes


• Biological processes
• Transport processes

Figure 1. Contaminant-related processes in soil and ground water.

Some individual processes will be described and discussed in the following sections.

4.1. Physical and Chemical Reactions Processes

Physical and chemical reactions operate on the surface and in the subsurface by transferring
mass among the fluids, gases, and solids, and changing the chemical form of a solute.
Important physical and chemical processes in the subsurface include: 1. the dissolution of
immiscible liquids into water, 2. the dissolution/ precipitation of solids, 3. gaseous dissolution
and exsolution, 4. volatilization, 5. chemical reactions, and 6. sorption.
Chemical dissolution of immiscible liquids or solids into water is governed by the
compound’s solubility, which is an empirically observed property defined, at any given
temperature, as the compound's aqueous concentration when at equilibrium with an excess of
an immiscible liquid or a solid (given sufficient time to exchange). The process of chemical
precipitation is defined as the emergence of solids from a solution, potentially occurring when
the solution is supersaturated-when the water is holding more dissolved solutes than its
solubility allows. The rate of dissolution or precipitation is related to the speed of
groundwater movement, and to chemical changes within the water. For example, in swift
flowing groundwater equilibrium will not be reached because the dissolved chemicals will be
174 Z. Yu, Y. Huang, A. Baron et al.

carried away and dissolution will therefore continue at increased rates compared to a stagnant
system. The aqueous solubilities of various organic compounds are available in the literature
(Domenico and Schwartz, 1990). When the solubility for a particular solute is not available, it
can be estimated through various methods (Lyman et al., 1981) such as the measurement of
the octanol/water partition coefficient in the laboratory, or from molecular structure.
Gas dissolution and exsolution can transfer significant quantities of mass between soil
gases and pore water in the subsurface. Henry's law is usually used to estimate distribution
between the two phases:

P aq
KH = (8)
P gas

where KH is the Henry's law constant (M/atm), Paq is the molar concentration of gas in the
solution (M), and Pgas is the partial pressure of gas in the gas phase (atm).
Volatilization is a process of either liquid evaporation or sublimation of solids to the
vapor phase. Volatilization of organic contaminants in unsaturated zones changes their
concentration and movement in the subsurface, and can cause vapor release to the atmosphere
(Fig. 1). The volatilization of dissolved organic solutes from water is described by Henry's
law, while volatilization of pure solvents is described by the liquid's vapor pressure.
Individual components of NAPL mixtures will have less volatilization into the gaseous phase
than those same components existing as individual, pure, and unmixed NAPLs, according to
Raoult's Law, which states that the mole fraction of the individual organic liquid in a mixture
equals the ratio of its partial pressure in the gas phase to the potential vapor pressure of the
pure organic solvent in a closed container. Because volatilization is typically a process in
which a liquid seeks to reach equilibrium with its overlying gaseous phase, any gaseous
movement that replaces the air in the unsaturated subsurface will enhance the rate of
volatilization.
Sorption between solutes and the surface of solids plays an important role in the fate and
transport of contaminants. On the one hand, contaminants that can attach easily to soil organic
matter and grain particles will stay near the places where they were initially placed or applied.
A limited amount of contaminants will enter groundwater by leaching through the macropores
or cracks and other preferential flow paths. The combined processes of overland flow and
erosion may carry these contaminants quickly to nearby streams. On the other hand,
contaminants that can dissolve easily in water and have low sorption will make their way
quickly to groundwater. Most contaminants exist in both sorbed and dissolved phases.
Usually, sorption isotherms are used to describe how mass partitions between the solution and
the solid when the solution is mixed with a solid medium. Sorption isotherms can be
described by expression of the general form:

S = f ( K d , C) (9)

where Kd is a partition coefficient reflecting the extent of sorption, C is the equilibrium


concentration (mg/L), and the function f which refers them depends on the particular solution
and solid. A linear expression is the simplest sorption isotherm (S=KdC). Real isotherms can
be linear, concave, convex or a combination of all three. In practice, sorption is modeled by
Contaminants in Groundwater and the Subsurface 175

fitting an experimentally derived isotherm to one of many theoretical equations. For the
hydrophobic sorption of organic compounds, Kd is calculated through the fraction of organic
compounds in soil and the coefficient of partition of a compound between organic carbon and
water (Koc). For many organic solutes, the amount of sorption is directly proportional to the
fraction of organic compounds in the soil. Various regression equations are available for
describing the relationship between Koc and the octanol-water partition coefficient or
solubility (Karickhoff et al., 1979; Schwarzenbach and Westall, 1981; Hassett et al., 1983).
There are many other chemical reactions that are not covered here and are also important
in problems of contamination of natural water. For example, acid-base reactions influence the
pH and the ion chemistry in water, which control other chemical processes.
Oxidation-reduction (Redox) reactions can change the solutes and chemical species
themselves in aqueous solutions, which in turn can affect sorption, mobility, and physical
processes such as volatilization. As mentioned previously, complexation reactions are an
important factor in the chemistry of aqueous systems (e.g., facilitating the transport of
potential toxic metals such as cadmium, chromium, and lead).

4.2. Biological Processes

Contaminants in the soil and groundwater will enter the biological system of plants,
consequently undergoing the full range of biological fate and transport mechanisms. Their
associated biodegraded components can accumulate in different parts of plants. It is very
important to understand these processes because animals and humans are directly exposed
when they eat contaminated plants.
In addition, the microorganisms (mainly bacteria) that occur ubiquitously in the
subsurface can play an important role in the prevention and remediation of some organic
contaminations of the soil and groundwater. These microorganisms mediate redox reactions
and use these reactions as a source of energy. Redox reactions control the speciation and
mobility of metal ions in a solution. More importantly, they are a major factor in
biodegradation, the biologically catalyzed biotransformation of organic compounds into
simpler inorganic forms such as carbon dioxides and water. Although metabolites, or
daughter products, produced by biodegradation and biotransformation sometimes offer little
or no environmental hazard, They often have similar or even greater environmental effects
than the parent compounds. Accordingly, understanding biotransformation and
biodegradation, how they convert parent compounds into various metabolites, and the
mobility and toxicity characteristics of these metabolites, is very important.

4.3. Mass Transport

Mass transport in soil and groundwater follows the mass conservation law. Solute transport is
conducted through the combination of advection, diffusion, and dispersion. Advection is the
movement of dissolved solutes with water flow. Advective flux can be expressed as:

J=qC (10)
176 Z. Yu, Y. Huang, A. Baron et al.

where J is mass flux, C is contaminant concentration, and q is the volumetric darcian flux of
water.
Because of the natural Brownian motion of dissolved ions and molecules, mass transport
in fluids can also occur via diffusion. A form of Fick's law of diffusion can be written as
follows:

∂C
J d = - θ Dd (11)
∂z

where Dd is the diffusion coefficient of a porous medium.


Mixing during mass transport is controlled by the combination of diffusion and
mechanical dispersion. The mechanical dispersion is caused by the variation in the field of
local velocity and can be described by an equation similar to Equation 11, but with a
mechanical dispersion coefficient (Dm). In some cases, the mechanical dispersion is scale-
dependent and more important than diffusion. Because Dd and Dm are similar, the coefficient
of hydrodynamic dispersion (D=Dd +Dm) is used to account for the combined effect of
diffusion and mechanical dispersion (Domenico and Schwartz, 1990). A general form of mass
transport in one dimension can then be described as:

∂( θC) ∂ ∂C
= [( θ D ) - q C] (12)
∂t ∂z ∂z

This equation can be further generalized to incorporate terms for chemical reactions and
sorption effects. Various numerical schemes (e.g., finite difference and finite element
methods) are available for solving for the temporal and spatial distribution of contamination
in one, two, and three dimensions.

5. Detecting and Monitoring Contaminant Movement


After potentially contaminated sites are identified, preliminary monitoring-related activities
are required in order to detect and monitor contaminant movement. An understanding of
surface soil and subsurface geology and their effect on storage and fluid transmission helps
determine the type of transport mechanisms (Kramer and Keller, 1995) at work. Geologic
material and the interaction between its variants vary, of course, from site to site, can be
determined through site visits, geologic maps, and subsurface investigation (e.g., drilling).
Various hydraulic parameters can be estimated through field experiments and laboratory
analysis. In addition to defining flow conditions, site characterization also typically includes
an estimation of the mass balance and phase transfer of contamination. Durant and Myers
(1995) described various approaches used by the United States Environmental Protection
Agency (EPA) to monitor the vadose zone at RCRA facilities, and the EPA's National
Exposure Research Laboratory has produced a useful CD-ROM entitled "Site
Characterization Library" (EPA/600/C-98/001, April 1998). A general, phased strategy for
site investigations begins with noninvasive techniques, methods that do not appreciably
disrupt the subsurface environment or cause a change in contaminant transport properties.
Contaminants in Groundwater and the Subsurface 177

They can be followed with more invasive techniques, which are selectively used and
optimized based on the results of the noninvasive techniques.

5.1. Noninvasive Techniques

Because contaminant movement in the subsurface can be accelerated by drilling activities and
other perturbations, many environmental agencies recommend beginning a site investigation
with techniques either conducted on the land surface or which do not require deep penetration
of the site. These techniques include: survey of records associated with the site, analysis of
past site photography and aerial photography, geophysical methods, and soil gas surveys.
Key site information pertaining to contaminant type, amount, distribution, and source can
often be found in records, manifests, historical maps, and regulatory compliance documents.
Records and interviews of personnel who have worked at the location can provide helpful
information, ranging from anecdotal comments to hard data. Other information that should be
reviewed includes: changes in site ownership and industrial processes, types of and quantities
of chemicals used in industrial processes, past location of storage facilities, pipelines and
loading docks and even weather data. An initial site inspection can take advantage of this
information and provide crucial clues in identifying pollutants and source areas. Past site
photographs and aerial photography of a site can also produce incisive spatial and temporal
data, showing exact times and location of leaks as well as geologic fault lineaments. There are
many sources for aerial photographs, including the U.S. Geological Survey's air photography
inventory at the Earth Science Information Center in Reston, Virginia.
There are numerous geophysical methodologies that remotely examine the subsurface
and produce information relating to the geological structure, location of tanks or cavities,
water table elevations, soil moisture content, and other subsurface parameters. These methods
can be applied from the ground surface or in boreholes drilled at the site, with the former
being noninvasive. Surface geophysical techniques include electromagnetic surveys,
resistivity, magnetometry, seismic methods, and ground penetrating radar. Other low impact
methods include time-domain reflectometry and tensiometry, and more intrusive borehole
methods include self-potential, natural gamma, gamma-gamma, neutron logs, temperature
profiles, downhole video, flowmeter measurements, and caliper logs.
Soil gas surveys are based on the principle that high concentrations of volatile
contaminants in the subsurface will be evident from elevated soil vapor concentrations of
those contaminants or their volatile daughter products. Measurement of soil gases involves
either active aspiration of soil by pumping or passive collection of adsorbed soil gases by a
buried device.

5.2. Soil Sampling

Soil sampling is important in assessing contaminant transport in the unsaturated zone and
locating the source of environmental spills or leaks. It is a requisite for understanding site
lithology and hydrogeology, estimating hydraulic parameters through geotechnical testing,
extracting microorganisms, and performing pore-liquid and soil chemical analyses. Soil
samples can be obtained using various sampling devices (Dorrance at al., 1995). Bulk
178 Z. Yu, Y. Huang, A. Baron et al.

samplers such as spades, shovels, scoops, and knives can be used to collect disturbed soil
samples. Auger samplers, which include screw-type augers and barrel augers, are readily
available and widely used because they are relatively inexpensive and easily operated.
However, they cannot be used at great depths or for stony soil. In some cases, auger sampling
can cause cross-contamination and samples obtained with augers may be disturbed. Tube
samplers can be used for obtaining fairly undisturbed soil cores up to 7 meters in depth in
unconsolidated sediments, but are generally more expensive and more difficult to operate.
Quality assurance procedures should be carefully followed. In practice, soil samples should
be stored in controlled-temperature refrigerators until analysis and prepared for analyses
using standard procedures.
Because many contaminants are colorless, it is not always obvious which soil samples are
contaminated. The result is that many samples are often unnecessarily sent to the laboratory,
which is not cost-effective. There are several quick field screening techniques for identifying
contaminated soil types. They are, specifically, the use of immunological surveys,
fluorescence techniques, centrifugation of liquids, and hydrophobic dye shaker tests.
Immunological surveys are colorimetric tests that measure the response of polyclonal
antibodies to certain types of pollution, while fluorescence methods take advantage of the
natural fluorescence of some contaminants (such as petroleum products) under black lights.
Another method for determining whether a clear liquid in the soil is water or a NAPL
involves centrifugation of the liquids, sometimes requiring individual droplets to be
withdrawn via syringe and placed in water. Some NAPLs can be distinguished from water by
mixing a hydrophobic dye and a soil sample in a plastic bag. The NAPL will preferentially
combine with the dye, indicating pollution.

5.3. Groundwater Sampling

Groundwater samples are needed to delineate and monitor contaminated plumes. A carefully
designed network of wells and piezometers is of major importance for sampling. Its design
must take into account the character and complexity of flow (e.g., flow direction and
hydraulic gradient). To accurately delineate the boundaries of the contaminant plume, wells
should be installed within and closely adjacent to the plume (Domenico and Schwartz, 1990).
In addition to direct water quality sampling, wells can be used to determine the direction of
groundwater flow or to conduct aquifer tests that estimate the magnitude of groundwater flow
parameters.
The most common sampling method involves purging a well at a high rate and then
obtaining the required volume of water, although low-flow purging can avoid disturbance of
the sampling point and spreading of contaminants (EPA, 1995). Samples help determine the
overall length, width, and thickness of the plume. Quality assurance procedures are needed to
assess the accuracy of the chemical data to avoid potential problems such as sloppy field and
laboratory practices, contamination of samples with fluids used for drilling, and sample
deterioration. There exist various multilevel sampling devices for obtaining samples of
solutes and nonaqueous phase liquids (NAPLs) in cased or uncased boreholes in the zone of
saturation. As mentioned earlier, geophysical techniques such as electromagnetic conductivity
methods have been used to determine plume boundaries. Because of the uncertainty of data
conversion and interpretation in later methods, they must be combined with a direct sampling
Contaminants in Groundwater and the Subsurface 179

procedure to ensure the definition of a plume. Newer methods involve circulating a suite of
partitioning tracers in the subsurface to measure groundwater flow and the mass of a stagnant
immiscible liquid by observing the differential tracer retardation. Detailed information on
subsurface sampling methods and general sample design can be found in EPA (1991).

6. Modeling Contaminant Transport in the Soil and Groundwater


Modeling contaminant transport based on the field studies provides an alternative approach.
Models can serve as platforms and repositories for new knowledge. They compile, integrate
and analyze the spatial and temporal parameters and variables that describe the complexity of
subsurface flow and contaminant transport (Anderson et al., 1993). Many models have been
developed and many modeling-related studies have been conducted in the last 30 years. The
results of these studies have improved our knowledge and understanding of contaminant
transport in the soil and groundwater, but there have unfortunately been many cases of
misuse, in which rigorous modeling practices were not followed.
Anderson and Woessner (1992) provide an eight-step modeling protocol. Before any
model is developed or selected, the purpose of the model must be clearly defined. Based on
defined objectives and the surface and subsurface setting of the study area, one can construct
a conceptual model. The conceptual model then provides a basis for selecting a mathematical
model from a suite of available public and commercial model packages, or, in some cases, for
designing a new model to fit specific requirements. Most of the available public and
commercial model packages have been verified against analytical solutions. Setting up the
model for a real simulation must take into account how well the site-specific contamination
problems are understood and how much data are available. This is an iterative process.
Sometimes, further field and laboratory work is required to collect key hydraulic parameters.
The next step, calibrating the model, requires calibration targets (e.g., hydraulic head and
contaminant concentration). Various objective techniques for model calibration and
uncertainty analysis are available, including Monte-Carlo (Beven and Binley, 1992) and latin
hypercube approaches (Yu and Schwartz, 1999) that utilize the plausible range of parameters.
Because of the computational burden of each simulation run, trial-and-error calibration is
usually favored over the more exhaustive calibrations that require many model runs with
different combinations of parameters. The values of soil and groundwater hydraulic
parameters (e.g., conductivity, retardation factor) are optimized to obtain a good fit between
the simulated variables (runoff and baseflow) and the observed data.
Following model calibration, model verification and predictive simulation must be
conducted to evaluate how the model responds to various possible stresses. Observational
data are essential to test and confirm the predictive capability of the model for research
purposes and regulatory and management applications. The final step is to summarize all
modeling results, including modeling performance and uncertainty analyses.
There are numerous numerical models that have been developed over the years to study
the fate and transport of contaminants in both unsaturated and saturated zones. In these
models, various finite-difference or finite-element numerical solution schemes are routinely
used to solve the governing equations described in previous sections. For example,
FEMWATER (Yeh, 1987), FEMWASTE (Yeh and Ward, 1981), and their derivatives use a
finite-element procedure and can be used to simulate water flow and contaminant transport in
180 Z. Yu, Y. Huang, A. Baron et al.

unsaturated and saturated zones. The following sections provide a general evaluation of
models of contaminant fate and transport in the soil and groundwater.

6.1. In Soil

SESOIL, PRZM, and PESTAN are three widely used models of contaminant fate and
transport developed by the EPA. SESOIL stands for the Seasonal Soil Compartment model
and was developed to simulate water flow, sediment transport, and contaminant fate in the
unsaturated zone (Mill et al., 1982). It incorporates infiltration, runoff, evapotranspiration,
percolation, and other hydrologic components with advection and dispersion of
contamination, sorption, volatilization, degradation, and other processes described in previous
sections and in Figure 2.

Figure 2. Process diagrams for various remediation schemes.

PRZM is a pesticide root zone model (Carsel et al., 1984) developed to simulate
contaminant movement within and below the plant root zone. Major hydrologic components
such as runoff, erosion, and evapotranspiration are simulated based on the curve number
techniques of the Soil Conservation Survey of the United States Department of Agriculture
and the water balance approach. The model includes advection, dispersion, and degradation,
but unlike SESOIL and PESTAN, does not include vapor-phase partitioning and transport.
PESTAN was developed for estimating the movement of organic contaminants through soil
to groundwater. Linear or nonlinear (Freundlich) sorption and first-order decay are
implemented in the model. Key parameters such as groundwater recharge rate, decay rate, and
sorption coefficient require sensitivity analysis. Donigian and Rao (1986) conducted a brief
comparison of the three models. All three were also evaluated through laboratory column
leaching experiments by Melancon et al. (1986). Simulated results (Jones et al., 1986) using
PRZM compare well with the field data.
Contaminants in Groundwater and the Subsurface 181

6.2. In Groundwater

Anderson et al. (1993) described a suite of available analytic and numerical flow and
contaminant transport models. While there are many codes for contaminant transport
simulations in the groundwater, the few described here are most widely used. These models
solve the advection-dispersion equations numerically in one, two, or three dimensions.
BIO1D, a simple physical model, can be used to simulate one-dimensional dispersion,
linear or nonlinear sorption and degradation at a constant velocity field (Srinivasan and
Mercer, 1987). With the input of initial boundary conditions and other parameters such as
velocity, longitudinal dispersion coefficient, and source concentration, two finite-difference
equations are solved for solute concentration. RNDWALK (Prickett et al., 1981), a
random-walk model, and MOC from the United States Geological Survey (Konikow, L.F.,
and Bredehoeft, J.D., 1978) were developed to simulate transient contaminant movement in
two dimensions. These models also include a code for solving two-dimensional hydraulic
heads and use derived hydraulic heads for computing velocity fields. Because a mass is
assigned to each particle, concentration need not be calculated at each time step but a high
density of particles is required for accurate solutions. Due to the random-walk approach,
solutions vary between successive runs and may not converge in some cases. Three-
dimensional versions are now available for these models. MT3D (Zheng, 1990) was
developed to simulate three-dimensional solute transport without considering heat and density
effects while SWIFT/386 (Ward, 1991) was developed to simulate variable-density flow, heat
and solute transport, and radionuclides. MT3D includes three-dimensional dispersion and
chemical reactions and various solution procedures.

7. Contaminant Remediation
The previous sections described major hydrologic processes in the soil and groundwater,
discussed flow and contaminant transport (including chemical reactions) in unsaturated and
saturated zones, introduced various approaches to detecting and monitoring contamination,
and summarized a number of available codes for modeling contaminant transport in the soil
and groundwater. Various approaches to remediate contaminated sites are discussed in this
section (Figure 2).
The most straightforward method is source containment. Physical prevention of any
further migration of contaminants is accomplished using slurry walls, sheet pile, grouting, and
geomembrances (OTA, 1984; Domenico and Schwartz, 1990). Slurry walls and sheet pile can
be easily installed in the soil and in soft subsurface materials, while grouting is efficient and
effective in a fractured rock media. Problems can arise when hydraulic gradients build up
across the physical barrier. Surface seals and drainage controls are used to avoid any
rainfall-induced infiltration in the contained contaminated sites, which can enhance flow and
movement of contaminants. Surface runoff can carry contaminated soil grains into nearby
streams or reinfiltrate contaminated water along flow paths. In general, surface drainage
control is designed to avoid any significant amount of topographically controlled
surface-runoff contribution to the contaminated sites. While source isolation using physical
barriers is helpful, it alone does nothing to remove or reduce subsurface contaminant mass.
182 Z. Yu, Y. Huang, A. Baron et al.

Excavation, fixation, and soil venting are available techniques for removing contaminants
in the unsaturated zone and soil. Generally, for small spills, excavation and removal is the
most cost effective solution. In these cases, contaminated materials can be excavated and
either treated or stored in other safe places. Fixation and encapsulation methods can be used
to reduce contaminant leaching. Soil venting is only used to remove volatile organic
compounds from the unsaturated zone and is based on the principle of enhancing
volatilization by the aspirating gases to create subsurface vapor circulation.
When contaminants enter the groundwater, the problems become more complicated.
Interceptor systems such as drains and trenches are used to collect contaminants close to the
water table. Pump-and-treat systems have been widely used to remediate various
contaminated sites, but, because of their inefficiency in mass removal, they are currently used
largely as hydraulic barriers. When combined with physical barriers, these systems can
extract contaminated groundwater from the subsurface and intercept aqueous contaminant
plumes. The extracted contaminated groundwater must then either be treated or directed to
nearby streams (if it will not affect the stream water quality significantly). Air stripping can
be used to remove the volatile contaminants in the extracted contaminated groundwater,
granular activated carbon can be used to remove dissolved organic contaminants, and
biological systems can be used to remove biodegradable contaminants.
Because of the nature of various contaminants (e.g. NAPLs) and their associated
chemical reactions and physical processes, many of them cannot be extracted quickly.
Contaminated water can be removed and replaced with clean water, but undissolved
contaminants remaining in subsurface media in the saturated zone will slowly re-contaminate
the groundwater. The result is that very long periods of time are required for pump-and-treat
to be effective and for contaminated water in the saturated zone to reach a required drinking
water standard or other cleanup goal. Complete restoration of the groundwater is often
impossible with pump-and-treat. The limitations of the system were not fully realized until
1989 (NRC, 1994). Due to our poor understanding of contaminant-related processes, the
majority of sites that used only pump-and-treat systems have not been restored to drinking
water standards. The feasibility of contamination cleanup with pump-and-treat systems
depends on site-specific contaminant chemistry, site geology, and the quantity and duration of
contamination. In light of these factors and the long cleanup times required, pump-and-treat
systems alone are often impractical for restoring the groundwater quality. Alternatives are
required.
Various enhanced pump-and-treat and alternative technologies have been developed and
used in the last decade. Pulsed and variable groundwater pumping allows the contaminant
concentration to build up so that contaminants can be removed with higher efficiency. Many
NAPL contamination problems are dealt with by directly pumping the nonaqueous phase,
which tremendously increases mass removal. For LNAPLs, this requires installation of
floating skimmer system pumps in recovery wells. Soil vapor extraction can be coupled with
groundwater pumping to extract organic contaminants by flushing air in the unsaturated zone
near a created cone of depression. Similar to soil vapor extraction, air sparging uses circulated
air to remove volatile contaminants, although sparing is carried out in the saturated zone, and
involves injection of air, not solely removal. The air stream carrying contaminants from a
sparged saturated zone must be captured by well-designed soil vapor extraction systems. In
situ bioremediation systems inject stimuli (e.g., oxygen) to stimulate subsurface
microorganisms (e.g., bacteria) to biodegrade hydrocarbon contaminants in the saturated zone
Contaminants in Groundwater and the Subsurface 183

into innocuous products, such as carbon dioxide and water. Bioventing is an in situ
bioremediation method for treating petroleum hydrocarbons and some chlorinated solvents in
the unsaturated zone. Like soil vapor extraction, it provides airflow in the system, but
whereas soil vapor extraction simply tries to volatilize contaminants the purpose of
bioventing is to supply oxygen (or inorganic nutrients) to microbes to enhance aerobic
biodegradation. There are variations of these approaches, such as bioslurping, which
concurrently aspirates water, LNAPL, and soil vapor from near the water table. Another
method is soil flushing, which can be used to wash sorbed contaminants into the groundwater
and enhance contaminant recovery in traditional pump-and-treat systems. Technologies in
development include in situ chemical treatment (e.g. surfactant flooding, oxidation), in situ
thermal techniques, steam enhanced extraction, and radio frequency heating.
There are alternative approaches that use other methods to attenuate the possible hazards.
Intrinsic bioremediation, for example, allows native microbes in the natural system to
biodegrade contaminants without human stimulation. In November of 1997 the U.S.
Environmental Protection Agency's Office of Solid Waste and Emergency Response issued
Directive 9200.4-17, guidelines for monitored natural attenuation practices. Monitored natural
attenuation is not a default or presumptive remedy, nor is it a no-action or walk-away
approach. The Directive prescribes source control and performance monitoring as
fundamental elements, and is applicable to all Superfund, RCRA, and underground storage
tank sites. A second alternative approach that is gaining acceptance is the use of in situ
reactive barriers that treat the contaminant plume by reducing contaminants as they pass
through subsurface permeable walls. Permeable barriers are often used in conjunction with
impermeable barriers that funnel groundwater toward the treatment zone. They must be
carefully monitored and maintained.
An integrated remediation system that combines various technologies is frequently
necessary to effectively and efficiently remediate contaminated sites and restore groundwater
to appropriate cleanup levels. In complex geological situations or where multiple
contaminants exist, treatment trains, which use one method after another, and/or phased
remedial approaches are often necessary. Modeling can be used to help evaluate various
alternatives and help design the remediation systems, particularly in situations of great spatial
heterogeneity.

8. Summary
Remediation of groundwater contamination has become increasingly important as
groundwater use expands. Subsurface contaminants include microbial pathogens, and
inorganic and organic chemicals, each of which has different abilities to migrate and
transform in the subsurface. Specific chemical contaminants can be toxic, carcinogenic,
teratogenic, or cause endocrine disruption, and are therefore important health considerations.
Fundamental hydrologic processes in the soil and groundwater are crucial to understanding,
describing, and predicting subsurface contaminant motion. At a particular field site, the
subsurface fate and transport of contaminants can be monitored through field sampling and
characterization efforts. The various field practices which help achieve this understanding
initially center on noninvasive measurements, such as analysis of aerial photography, surface
184 Z. Yu, Y. Huang, A. Baron et al.

geophysical techniques, and soil gas measurements, where applicable. Soil and groundwater
then can be sampled. In this phase, monitoring wells may be drilled and soil cores retrieved.
Mathematical simulation and prediction can be an important tool, allowing for a better
understanding of processes affecting contaminant transport and providing estimates for
regulatory, management, and remediation purposes. There are many models that are used for
simulating flow and contaminant transport in unsaturated and saturated zones.
Various techniques are available for remediating contaminated sites, and new
technologies are presently under development. The feasibility of any remediation technology
should be carefully evaluated from site to site, based on site-specific information such as local
geology and type of contamination. In some cases, a combination of treatment methods, used
in a phased approach, is more efficient and effective than one method alone. With the number
of known contaminated sites continuing to increase, there is an urgent need for scientists and
regulatory agencies to conduct more fundamental research to gain a better understanding of
the physical processes behind the contamination and to develop more innovative remediation
methods for the soil and groundwater.

Acknowledgments

This project is partially supported by Virgin Valley Water District and Program for
Changjiang Scholars and Innovative Research Team in University (PCSIRT)(IRT0717).

References
Anderson, M.P., and Woessner, W.W., 1992. Applied Groundwater Modeling: Simulation of
Flow and Advective Transport, Academic Press, Orlando, Fla., 379 p.
Anderson, M.P., Ward, D.S., Lappala, E.G., and Prickett, T.A., 1993. Chapter 22: Computer
models for subsurface water. In Handbook of Hydrology (ed. M.R. Maidment), McGraw-
Hill,
Beven, K., and Kirkby, M.J., 1979. A physically based, variable contributing area model of
basin hydrology. Hydrol. Sci. Bull., 24: 43-69.
Beven, K., 1989. Changing ideas in hydrology-the case of physically-based models. J.
Hydrol., 105: 157-172.
Beven, K., and Binley, A.M., 1992. The future of distributed models: model calibration and
uncertainty prediction. Hydrol. Process, 6: 279-298.
Carsel, R.F., Smith C.N., Mulkey, L.A., Dean, J.D., and Jowise, P., 1984. Users=s Manual
for the pesticide root zone model (PRZM). EPA-600/3-84-109, Athens, GA.
Clapp, R.B., and Hornberger, G.M., 1978. Empirical equation for some soil hydraulic
properties. Water Resour. Res., 14(4): 601-604.
Cosby, B.J., Hornberger, G.M., Clapp, R.B., Ginn, T.R., 1984. A statistical exploration of the
relationship of soil moisture characteristics to the physical properties of soils. Water
Resour. Res., 20: 682-690.
Davis, S.N., 1969. Porosity and permeability of natural materials. In flow through porous
material (ed. R.J.M. DeWiest). Academic Press. p. 54-89.
Contaminants in Groundwater and the Subsurface 185

Domenico, P.A., and Schwartz, F.W., 1990. Physical and chemical hydrogeology. John Wiley
& Sons. 824 p.
Donigian, A.S. Jr., and Rao, P.S.C., 1986. Overview of terrestrial processes and modeling. In
Vadose Zone Modeling of Organic Pollutants (eds. S.C. Hern, S.M. Melancon), Lewis
Publishers, Chelsea, Michigan.
Dorrance, D.W., Wilson, L.G., Everett, L.G., and Cullen, S.J., 1995. A compendium of soil
samples from the vadose zone. In Handbook of Vadose Zone Characterization &
Monitoring (eds. L.G. Wilson, L.G. Everett, S.J. Cullen), Lewis, p. 401-428.
Dunne, T., 1978. Field studies of hillslope flow processes. Hillslope Hydrology, M. Kirkby,
Ed. Wiley. p. 227-294.
Dunne, T., and Black, R.D., 1970. Partial area contributions to storm runoff in a small New
England watershed. Water Resour. Res., 6: 1296-1311.
Durant, N.D., and Myers, V.B., 1995. EPA=s approach to vadose zone monitoring at RCRA
facilities. In Handbook of Vadose Zone Characterization & Monitoring (eds. L.G.
Wilson, L.G. Everett, S.J. Cullen), Lewis. p. 9-22.
Environmental Protection Agency (EPA), 1991. Site characterization for subsurface
remediation, Office of Research and Development, Washington, 259 p.
Environmental Protection Agency (EPA), 1995. Ground water sampling - a workshop
summary, Office of Research and Development, Washington, 98 p.
Freeze, R.A., 1974. Streamflow generation. Water Resour. Res., 12: 627-647.
Freeze, R.A., 1980. A stochastic-conceptual analysis of rainfall-runoff processes on a
hillslope. Water Resour. Res., 16: 391-408.
Green, W.H., and Ampt, A.A., 1911. Studies on soil physics, part I, the flow of air and water
through soils, J. Agric Sci., vol. 4, no. 1, p. 1-24.
Hassett, J.J., Banwart, W.L., and Griffin, R.A., 1983. Correlation of compound propoties with
sorption characteristics of nonpolar compounds by soils and sediments: Concepts and
limitations. In Environment and Solid Waste: Characterization, Treatment and Disposal
(eds. C.W. Francis and S.I. Auerbach, Stoneham, Mass., Butterworth), p. 161-178.
Horton, R.E., 1933. The role of infiltration in the hydrologic cycle. Trans. Ameri. Geophys.
Union, 14: 446-460.
Junes, R.L., Black, G.W., and Estes, T.L., 1986. Comparison of computer model predictions
with unsaturated zone field data for aldicarb and aldoxycarb. Environ. Toxicol. Chem., 5,
p. 1027-1037.
Karickhoff, S.W., 1981. Semi-empirical estimation of sorption of hydrophobic pollutants on
natural sediments and soils. Chemosphere, 10(8): 833-846.
Konikow, L.F., and Bredehoeft, J.D., 1978. Computer model of two dimensional solute
transport and dispersion in ground water, In Techniques of Water-Resources
Investigations (Book 7), U.S. Geological Survey.
Kramer, J.H., and Keller, B., 1995. Understanding the geologic framework of the vadose zone
and its effect on storage and transmission of fluids. In Handbook of Vadose Zone
Characterization & Monitoring (eds. L.G. Wilson, L.G. Everett, S.J. Cullen), Lewis, p.
137-158.
Lyman, W.J., 1981. Solubility in water. In Handbook of Chemical Property Estimation
Methods (eds W.J. Lyman, W.F. Reehl, and D.H. Rosenblatt), McGraw-Hill Book
Company, chapter 2, p. 1-50.
186 Z. Yu, Y. Huang, A. Baron et al.

Melancon, S.M., Polard, J.E., and Hern, S.C., 1986. Evaluation of SESOIL, PRZM and
PESTAN in a laboratory column leaching experiment, Environ. Toxic. Chem., 5, 865-
878.
Mill, T., Mabey W.R., Bombwerger, D.C., Chou, T.W., Hendry, D.C., and Smith, J.H., 1982.
Laboratory protocols for evaluating the fate of organic chemicals in air and water. EPA-
600/3-82-022, Washington, DC.
National Research Council, 1994. Alternatives for ground water cleanup, National Academy
Press, Washington, DC., 315 p.
Office of Technology Assessment, 1984. Protecting the nation=s groundwater from
contamination. Office of Technology Assessment, OTA-0-233, Washington, D.C., 244 p.
Philip, J.R., 1957. The theory of infiltration: 1. The infiltration equation and its solution, Soil
Sci., vol 83, no. 5, p. 345-357.
Prickett, T.A., Naymik, T.G., and Lonnquist, C.G., 1981. A random-walk solute transport
model for selected groundwater quality evaluations, Ill. State Water Surv. Bull., vol. 65.
Rawls, W.J., and Brakensiek, D.L., 1985. Prediction of soil-water properties for hydrologic
modeling, in Watershed Management in the Eighties. Am. Soc. Of Civ. Eng., New York,
p. 293-299.
Schwarzenbach, R.P., and Westall, J., 1981. Transport of nonpolar organic compounds from
surface water to groundwater. Laboratory studies. Environ. Sci. Technol., 15: 1300-1367.
Srinivasan, P., and Mercer, J.W., 1987. BIO1D one dimensional model for comparison of
biodegradation and adsorption processes in contaminant transport, GeoTrans, Sterling,
VA.
Swartzendruber, D., 1969. The flow of water in unsaturated soils. Academic Press, New
York, p. 215-292.
van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic conductivity
of unsaturated soils. Soil Sci. Soc. AM. J., 44(5): 892-898.
Troendle, C.A. 1985. Variable source area models. In M.G. Anderson and T.P Burt, eds.,
Hydrological Forecasting, John Wiley & Sons, New York, 347-403.
Ward, D.S., 1991. Data input guid for SWIFT/386, version 2.50, GeoTrans Technical Report,
Sterling, Va.
Yeh, G.T., 1987. FEMWATER: a finite element model of water flow through saturated-
unsaturated porous media-first revision, ORNL-5567/R1, 248 pp., Oak Ridge Natl. Lab.,
Oak Ridge, Tenn.
Yeh, G.T., and Ward, D.S., 1981. REMWASTE: a finite-element model of waste transport
through saturated-unsaturated porous media, ORNL-5601, Oak Ridge Natl. Lab., Oak
Ridge, Tenn.
Yu, Z., and Schwartz, F.W., 1999. Automated calibration applied to constrained ground-water
flow modeling. Hydrol. Processes, 13: 191-209.
Zheng, C., 1990. MT3D, a modular three-dimensional transport model, S.S. Papadopulos and
Associates, Bethesda, Md.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 187-202 © 2008 Nova Science Publishers, Inc.

Chapter 7

PRELIMINARY STUDIES FOR DESIGNING A WETLAND


FOR ARSENIC TREATMENT

Raja Chowdhury, Defne Apul


Dept of Civil Engineering, University of Toledo, OH
Daryl Dwyer
Dept of Environmental Sciences, University of Toledo, OH

Abstract
Wetlands have been used successfully for treating domestic, pharmaceutical and mining
related wastewater. There has been recent interest in using constructed wetlands to treat
arsenic laden waste. In a wetland, arsenic may exist in either reduced or oxidized forms, as a
component of organic arsenic – containing species and as insoluble sulfide minerals. Research
has been done to characterize the fate and transport of arsenic in soils; however, little has been
done to characterize these mechanisms for wetlands, a step which is necessary in creating an
efficient wetland treatment system. The research presented in this paper describes the first
phase in modeling the fate of arsenic within a wetland and focuses on soil characterization and
estimation of the design life time using HYDRUS – 2D, a variably saturated flow and
contaminant transport model. A mixture of biosolids and sediments was selected as the soil to
be used in the proposed constructed wetland. The hydraulic conductivity and adsorption
coefficient for the soil were estimated using a soil – packed vertical column. Four different
scenarios were simulated: (i) top inlet and top outlet, (ii) top inlet and bottom outlet, (iii)
bottom inlet and top outlet, and (iv) bottom inlet top outlet with a gravel layer at the bottom.
The simulations were carried out for a bench scale (12 x 12x 54 cm) wetland but the results
were extrapolated to a field - scale wetland. Simulations showed that a design scenario of a
bottom inlet with top outlet (with or without gravel layer at the bottom) has the highest delay
for breakthrough of arsenic in the outlet. Other avenues for arsenic removal including plant
uptake, and arsenic speciation and precipitation were not included in the current simulation
due to limitations of the software, but are to be added as the research project continues.
188 Raja Chowdhury, Defne Apul and Daryl Dwyer

Arsenic as a Widespread Problem


Arsenic is one of the most prevalent heavy metals found in superfund and brown field sites
present in all over the USA (USEPA, 1997). Arsenic is also found in industrial wastewaters,
landfill leachate, and in mine waste. In Bangladesh, a considerable portion of cultivated land
is contaminated with arsenic due to use of (naturally occurring) arsenic laden groundwater for
irrigation and subsequent accumulation of arsenic in crops especially in rice (Liu et al., 2006).
In the USA, mining activities is the largest producer of arsenic containing waste. By volume
it is the fifth largest metal containing waste [http://www.pollutionissues.com/Te-Un/Toxic-
Release-Inventory.html].

Constructed Wetlands
Constructed wetlands have been used for many decades for treatment of municipal and
industrial wastewaters. In Europe and in the US, natural wetlands have been effectively used
for almost 100 years to treat municipal wastewater [Hiley, 2003; USEPA, 2000]. The waste
degradation in a wetland takes place by the organisms that are part of the ecosystem and
flourish in a wetland. Interest in constructed wetlands has grown significantly in the past
decades as a sustainable approach to treat wastewater [Crites et al., 2006]. Constructed
wetlands are considered eco – friendly because they don’t rely on industrial processes and
materials and they use natural attenuation processes to treat waste [ITRC, 2006].
Constructed wetlands have been used to treat metal laden wastewater on an experimental
basis [Ye et al., 2003; Buddhawong et al., 2005]. The cost of building and operating a
wetland is lower compared to other treatment strategies such as precipitation/co-precipitation,
adsorption, soil washing or conventional wastewater treatment plants [ITRC, 2003; EPA,
2002]. In a wetland, gravel/sand/soil are used, these are natural materials and need negligible
processes to produce the final products. In a conventional treatment, equipment and materials
used need an extensive production process and thus, during their production cycle, they
produce a heavy load of pollutants. Throughout their life cycle, wetlands consume less energy
and materials and they also support ecosystems. Hence, wetlands are a more sustainable, eco
– friendly solution for treating arsenic contaminated water.
Design life of a constructed wetland can vary significantly based on site conditions, the
treatment purpose and the specific mechanisms that work to remove pollutants in a wetland.
The mechanisms of arsenic removal using a constructed wetland are a juxtaposition of
adsorption, precipitation of metal sulfide and plant uptake. Among these mechanisms
adsorption and metals sulfide precipitation play a major role for arsenic removal [Cohen,
2006; Jong and Parry, 2003, 2004; Stein et al., 2007]. Metal sulfide precipitation takes place
in an anaerobic environment whereas adsorption of As (V) onto hydrous ferric oxides
becomes more favorable in an aerobic environment. Removal of arsenic is also affected by
the distribution of water and hydraulic retention time. Exhaustion of adsorption capacity,
inadequate microbial activities, exhaustion of buffering capacity, and clogging may reduce
the design life of wetlands constructed for removal of metal laden water. One way to estimate
the design life is to simulate the redox dependent chemistry of arsenic and its transport
through the system via advection, diffusion, and dispersion.
Preliminary Studies for Designing a Wetland for Arsenic Treatment 189

Software Platform
A literature review was done on available software for simulating the arsenic chemistry and
flow regime in wetlands. Visual MINTEQ is a powerful thermodynamic equilibrium model
for simulating the chemistry of arsenic. Visual MINTEQ is user friendly but it does not have
a flow modeling component (http://www.lwr.kth.se/english/OurSoftware/Vminteq/
index.htm). A recent model for water flow and plant uptake modeling was presented by
Verma et al. (2007). This model can simulate the pH dependent uptake of metals by plants.
However, the package can not simulate the redox dependent arsenic chemistry and microbes
mediated arsenic speciation. A few other software packages are also discussed in the literature
[Peake, 1996; Bar – Yosef et al. 2005; Ouyang, 2005]. However, none of these were suitable
for the purpose of the present work that required a capability to simultaneously handle arsenic
chemistry, plant uptake, and flow modeling. HYDRUS, is a popular commercial software that
can handle both flow and simple contaminant plant uptake calculations [Wohling and
Schmitz, 2007; Dousset et al., 2007]. HYDRUS can also be used to simulate flow of solutes
with retardation by considering, linear and nonlinear adsorption isotherm, ion exchange and
non equilibrium mass transfer. While HYDRUS is a powerful fate and transport model it can
not be used to model redox profile and resultant speciation of arsenic. For purposes of this
study, modeling of the fate and transport of arsenic in wetlands was reduced to modeling of
the flow of water (with dissolved arsenic) and retardation of arsenic by its adsorption onto
soil in oxic conditions. Arsenic breakthrough curve for the wetland system was simulated and
then using the breakthrough curve, the design life of a constructed wetland was estimated.

Research Objectives and Overview of Approach


The initial phase of the study was undertaken to: (i) select an appropriate soil for use in
constructing an experimental wetland and (ii) estimate fluid flow. In the next phase, the
experimental wetland was modeled to estimate (iii) the effects of various inlet outlet
distributions on transport of arsenic, (iv) the design life, and (v) the water budget.
Three types of soils were analyzed in the selection process: (a) Nusoil (32% clay, 65%
silt, 3% sand, and 4% organic matter content), which is a mixture of biosolids obtained from a
wastewater treatment plant and sediments dredged from the shipping channel for the port of
Toledo, Ohio located on Maumee Bay, Lake Erie; (b) dredged sediments alone (56% clay,
41% silt and 4% sand) and sediment (same as Nusoil but it has lower organic matter content).
The adsorption capacity of each soil was determined in batch experiments and the soil with
the highest adsorption capacity was selected for further experiments (column study) to
estimate both its hydraulic conductivity and dispersion coefficient. Adsorption coefficient
was used for modeling arsenic transport in the wetland along with other parameters
mentioned above. Using HYDRUS 2D the water budget was estimated to understand how the
precipitation and evaporation affect water inflow and outflow in the wetlands.
190 Raja Chowdhury, Defne Apul and Daryl Dwyer

Methods
Batch Adsorption Study

Adsorption is a mass transfer phenomenon between a gas/liquid phase and a solid phase.
Solutes present in a gas/ liquid phase can be adsorbed into the surface of the solid phase. The
adsorption phenomenon is often modeled using Langmuir (Eq. 1) and Freundlich isotherm
(Eq. 2) [Sawyer et al., 2003].

K ads C
q = qm (1)
1 + K ads C

q = KC n (2)

Where
q = adsorbed concentration, qm = maximum adsorption capacity, C = aqueous
concentration of solute, Kads = a constant, K = a constant, n = fitting parameter. When n = 1
the Freundlich model is reduced to a linear isotherm.
Soil samples were sieved through 2 mm sieve and then dried at 600 C for 2 days. Then,
soil samples were kept in room temperature for one day. From dried samples 10 gm soils
were weighed for the adsorption experiment. The adsorption experiment was carried out for
five different initial arsenic concentrations (i) 0.5 mg/L, (ii) 1 mg/L (iii) 5 mg/L (iv) 10 mg/L
and (v) 20 mg/L. Sodium nitrate (425 mg/L) dissolved in Millipore water was used to adjust
the ionic strength to 0.01 in the adsorption experiments.
Centrifuge tubes (Fisher, 3123 – 0250) consisted of 10 gm of soil mixed with 200 mL
water (ionic strength 0.01) with defined concentration of arsenic which was added from a
stock solution of arsenic (250 mg/L). The bottles were shaken for 2 days. The duration of the
experiment was determined from a kinetic study which showed that 24 hrs was needed for
equilibration. After shaking, the bottles were centrifuged at 3500 rpm for 30 min. Then, 10
mL was taken from supernatant and acidified by adding concentrated HNO3. Acid added was
3.5% by volume of 10 mL sample. Samples were analyzed by ICP – OES (optical emission
spectroscopy). Details about the ICP method is given in Rofkar et al. (2007).

Column Study

A 5 –cm diameter glass column was used for estimation of hydraulic conductivity and the dispersion
coefficient of the soil. Soil was sieved through a 2 mm sieve and then soil particles smaller than 2
mm were used for the study. 400 gm of soil was poured in the soil column and compacted. The final
depth of soil in the column was 19 cm. The soil was then saturated with water (tap water).
Bromide was used as a conservative tracer. 10 mL Br of 7,764 mg/L concentration was
poured on top of the soil column and allowed to enter into the soil column. After bromide
addition, a constant pressure head of 50 cm of water was applied over the soil column. Effluent
from the column was collected in every 48 min using an automated sampler. Samples were
collected for 51 hrs and Br concentration was analyzed using a bromide electrode. Volume of
each sample was also measured using a graduated cylinder.
Preliminary Studies for Designing a Wetland for Arsenic Treatment 191

Simulation Study

Estimation of Dispersion Coefficient Using HYDRUS 1D

Estimates of dispersion coefficient and hydraulic conductivity were made using HYDRUS 1
D. Measured concentrations of bromide and time of measurement were input in HYDRUS 1D
and HYDRUS 1D was run using the inverse solution option to simultaneously estimate the
dispersion coefficient and the hydraulic conductivity. The top and bottom boundary
conditions for flow were constant head and free drainage, respectively. Initially, in the
simulation, the column was completely saturated and the Br was input in the top 0.57 cm of
the column. The porosity used in the simulation was 0.45. Initial guesses of dispersivity and
saturated hydraulic conductivity were set to 0.6 cm2/hr and 0.25 cm/hr.

Simulation of Flow through a Wetland Using HYDRUS 2D

To save simulation time and computation effort, simulation studies were carried out in a
smaller prototype of the field scale wetland (20 ft x 4 ft x 4 ft). The dimension of the small
scale wetland (length = 54 cm, width and height = 12 cm) used in the HYDRUS 2D
simulation was 1165 times smaller (volume wise) than the field scale wetland. The hydraulic
conductivity, and the dispersion coefficient estimated from the column study and adsorption
data obtained from the adsorption experiments were used in the simulation (Table 1).

Table 1. Input parameters used in HYDRUS simulation

Water content (%) Hydraulic Bulk


Soil type conductivity density Adsorption coefficient
residual saturated (cm/day) (mg/cc)
K = 0.35 (cc/mg) (a), n = 0.8(a)
Nu soil 0.067** 0.45** 5(a) 1.6**
(from adsorption experiments)
Gravel 0.45* 0.43* 2000** 2* K = 0.05** n = 1**
a
** assumed, * from inbuilt data set of HYDRUS experimentally determined

Solute transport modeling was carried out in four different scenarios (i) top inlet and top
outlet (TOP – TOP) (ii) top inlet and bottom outlet (TOP – BOTTOM) (iii) bottom inlet and
top outlet (BOTTOM – TOP) and (iv) bottom inlet and top outlet with gravel layer at the
bottom (3 cm) (BOTTOM – TOP (GRAVEL). Simulations were carried out for bromide
(conservative tracer) and arsenic (reactive tracer). Other details about the model are as
follows (i) constant flux inlet at the left hand vertical wall (50 cm2/day); (ii) initially the soil
was saturated with water; (iii) atmospheric boundary condition was used for the surface of the
wetland; (iv) all other boundaries were no flux boundary condition, and (v) a seepage or
outlet at the right hand vertical wall was used. Simulations for conservative tracers (bromide)
were run for 40 days whereas simulations for nonconservative tracers (arsenic) were run for
100 days. The horizontal hydraulic conductivity is assumed to be 10 times higher than the
vertical hydraulic conductivity which was measured from column study.
192 Raja Chowdhury, Defne Apul and Daryl Dwyer

Water Budget

The dimensions of the wetland used in the HYDRUS 2D simulation for water budget
estimation were as follows: length = 610 cm (20 ft) and height = 122 cm (4 ft). The
simulation was carried out using a constant head boundary condition at the inlet which was
placed at the bottom left side of the mesh. For root water uptake the S shape model was used.
Root water uptake, precipitation and evaporation were included in the model (Table 2).
Volume of evaporation was based on information from monthly value available at
www.cpc.ncep.noaa.gov/soilmst/e.shtml. Transpiration data for wetland plants was collected
from Koch and Rawlik (1993). Precipitation data was obtained from NOAA (2004). The
simulation was carried out for 365 days.

Table 2. Precipitation evaporation and transpiration (cm) data used in the water budget
simulation

Time Precipitation Evaporation Transpiration


30 0.163 0 0.1
60 0.16 0 0.1
90 0.22 0.02 0.1
120 0.267 0.01 0.5
150 0.293 0.5 0.5
180 0.32 0.6 0.5
210 0.236 0.6 0.5
240 0.27 0.6 0.5
270 0.24 0.4 0.5
300 0.197 0.2 0.5
330 0.237 0.02 0.1
365 0.191 0 0.1

Inlet pressure head was 130 cm during the simulation period and initially the soil was
saturated with water. The simulation had the following boundary conditions: (i) atmospheric
boundary condition at top surface, (ii) no flux boundary condition at all other boundaries, (iii)
a constant head inlet (of 130 cm) at the left hand vertical wall, and (iv) a seepage face outlet
at the top of the right hand vertical wall. Cumulative inflow, outflow, evaporation and
transpiration were calculated by the software. Then dividing those values by the number of
days of simulation (365 days) the average daily inflow, outflow, evaporation and transpiration
rates were estimated.

Design Life

According to the environmental protection act of 2002, the discharge limit of arsenic in
surface water and in land/underground is 0.1 mg/L. Hence, the goal of the project was to
bring down the concentration of arsenic from 1 mg/L to 0.1 mg/L using wetland mediated
treatment. The design life of field scale wetland was estimated from the simulated
breakthrough curve of the small scale wetland. Adsorption of arsenate was assumed to be the
Preliminary Studies for Designing a Wetland for Arsenic Treatment 193

primary removal mechanism. Adsorption data from batch adsorption study was used for
simulation purposes. However, adsorption is typically affected by the liquid solid ratio; pH
and contact time [Haque et al., 2008]. Adsorption coefficient estimated from batch study can
be considered as the maximum possible value that can be reached. Yet, Haque et al., (2008)
found slightly higher adsorption coefficients in their column study compared to the batch
study. They also observed that due to increase in flow path (increase in column length)
adsorption capacity was also increased. Hence, it can be assumed that due to the increase in
flow path, adsorption capacity of a field scale wetland might be higher compared to a small
scale wetland. Hence, as a simplistic assumption, the design life of the field scale wetland was
estimated by multiplying the design life of a small scale wetland by the ratio of volume of
field scale wetland to the small scale wetland.

Results and Discussion


Adsorption Study

Data obtained from the batch experiments were fitted to Freundlich (linear as well as
nonlinear) and Langmuir isotherms. A perusal of the fit suggested that nonlinear Freundlich
isotherm described the data most accurately, followed by a linear isotherm model. Langmuir
model did not provide a good fit to data points (Figure 1). Hence, for comparison purposes,
only Freundlich isotherm parameters are discussed. Adsorption parameters obtained from
each type of plots are also summarized in Table 3.

0.54

0.45
q=0.351C^0.799

0.36 q = 0.3619C
q (mg/g)

K=0.362 L/g
0.27 2
R = 0.9685
0.18

0.09
Langmuir plot
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
equilibrium As conc. (mg/L)

(a)

Figure 1. Continued on next page.


194 Raja Chowdhury, Defne Apul and Daryl Dwyer

0.6
q=0.116 C^0.786
0.5
q = 0.1084C
0.4 K=0.1084 L/g
q (mg/g) R2 = 0.9909
0.3

0.2

0.1 Langm uir plot

0
0.0 1.0 2.0 3.0 4.0 5.0
Equlibrium As concentration (mg/L)

(b)
0.6
q=0.238C^0.687
0.5 q = 0.1636C
K=0.164 L/g
2
0.4 R = 0.9146
q (mg/g)

0.3
Langmuir plot
0.2

0.1

0
0 1 2 3
Equilibrium As Conc (mg/L)

(c)

Figure 1. Adsorption isotherm for (a) nu soil (b) dredged sediment and (c) sediment.

Table 3. Adsorption isotherm data for three kinds of soils

Type of soil Type of isotherm Isotherm parameters


Nu soil Langmuir qm = 0.148 mg/g Kads = 0.141 L/mg
Freundlich linear K=0.362 L/g
Freundlich nonlinear K= 0.351 L/g n= 0.8
Dredged sediment Langmuir qm = 0.692 mg/g Kads = 0.134 L/mg
Linear K= 0.108 L/g
Freundlich K= 0.116 L/g, n= 0.687
Sediment Linear K= 0.164 L/g
Langmuir qm = 0.319 mg/g Kads = 0.276 L/mg
Freundlich Kd=0.238, n=0.687
Preliminary Studies for Designing a Wetland for Arsenic Treatment 195

Nu soil had the highest adsorption capacity followed by sediment and dredged sediment
if one considers the K value of Freundlich isotherm (Table 3). At higher concentration of
arsenic (10 mg/L or higher initial concentration), all the adsorption isotherms showed some
nonlinearity.
A compilation of arsenic adsorption data for different media is available in Mohan and
Pittman (2007). A perusal of this data set shows that most of the adsorption data follows
Langmuir isotherm and some of them follow Freundlich isotherm. The adsorption
coefficients for different media vary from 0.008 – 15 L/g [Mohan and Pittman, 2007]. Three
kinds of soils we have tested had almost similar adsorption coefficient [K (L/g)] [0.116 –
0.35] and falls within the range found in literature.
The batch study was carried out to estimate which soil had better adsorption capacity for
arsenic. Arsenic concentration of the influent water was assumed to be around 1 mg/L. At that
initial concentration, the equilibrium concentration of arsenic (after adsorption) varies from
0.02 (sediment) to 0.09 (dredged sediment). For nusoil, at 1 mg/L of initial arsenic
concentration, the equilibrium concentration was 0.04 mg/L. Hence, the sediment and nusoil
had almost similar capacity to adsorb arsenic in the working range of the project. Their
adsorption coefficients (K) were also similar (Table 3). Hence, for the current project, nusoil
and sediment can be used as a medium. However, nusoil has higher organic matter content
which will be beneficial for microbial activities. Also nusoil is locally available and hence it
was selected as a more appropriate medium for the constructed wetland.

Column Study

The experimental observations and simulated column results for nusoil are shown in Figure 2.

1.20E+03

1.00E+03
Concentration Br (mg/L

8.00E+02
Conductivity: 0.499 cm/hr
2
Disp: 0.71 cm /hr
6.00E+02 Regression Coeff: 0.97

C
4.00E+02

2.00E+02

0.00E+00
0 10 20 30 40 50 60
Time (hr)

Figure 2. Experimental (scattered points) and modeled (continuous line) breakthrough curve for Br in
Nu soil.
196 Raja Chowdhury, Defne Apul and Daryl Dwyer

The peak of the breakthrough curve was observed around 18 hrs. The volume of water
collected at each 48 min intervals was around 6 – 8 mL. The hydraulic conductivity and
dispersion coefficient obtained from the HYDRUS 1D inverse simulation were 0.5 cm/hr and
0.71 cm2/hr, respectively.

Simulation of Flow through the Wetland

Breakthrough curve and velocity profile were modeled for four different inlet outlet
scenarios. As expected, the breakthrough time observed for bromide was lower compared to
the arsenic breakthrough time [Figure 3(a, b)]. Breakthrough time of arsenic is longer due to
adsorption of arsenic in soil. Breakthrough time for bromide and arsenic were around 25 days
and 75 – 95 days, respectively. The difference in breakthrough time was more prominent for
arsenic and it (breakthrough time) was longest for top inlet and top outlet condition. However,
the reason behind it is the loss of solute due to overflow. The overflow of water at top inlet
and top outlet condition was also apparent from the plot of velocity vectors [Figure 4(a)].
Adding a gravel layer did not affect the breakthrough time even though the adsorption in
gravel layer was almost 1/7 times of nusoil (Table 1). Adding a gravel layer did help to
distribute the flow in a better fashion than the other scenarios [Figure 4]. When gravel layer
was not added, the flow was mostly horizontal (except in few places); adding a gravel layer
caused the flow to be mostly in the vertical direction [Figure 4(d)].

(a) 6.00E-03

5.00E-03

4.00E-03
Conc. Br (mg/cc)

TOP TOP

3.00E-03
BOTTOM TOP

2.00E-03 TOP BOTTOM

1.00E-03 BOTTOM TOP (Gravel)

0.00E+00
0 5 10 15 20 25 30 35 40 45

Time (days)

Figure 3. Continued on next page.


Preliminary Studies for Designing a Wetland for Arsenic Treatment 197

(b)
6.00E-03

5.00E-03

4.00E-03
Conc As (mg/cc)

3.00E-03 BOTTOM TOP

TOP TOP
2.00E-03
TOP BOTTOM

1.00E-03 BOTTOM TOP


(Gravel)

0.00E+00
0 20 40 60 80 100 120 140
Time (day)

Figure 3. Breakthrough curve for various scenarios (a) for Br (b) for arsenic (Line shown at 0.001
mg/cc level is the desired breakthrough for the present case).

(a) Top inlet and top outlet

(b) Top inlet and bottom outlet

Figure 4. Continued on next page.


198 Raja Chowdhury, Defne Apul and Daryl Dwyer

(c) Bottom inlet and top outlet

(d) Bottom inlet and top outlet with gravel layer at the bottom

Figure 4. Velocity vectors in four different scenarios after 3-4 days. After 3-4 days, the magnitude and
direction of the velocity did not change much in all of the simulations.

When gravel layer was present, water first moved through the gravel layer (high
conductivity) and then moved upward towards the outlet. Mixing and distribution of water
was found maximum in TOP-TOP scenario [Figure 3 and Figure 4(a)] followed by
BOTTOM-TOP (GRAVEL) scenario [Figure 4 (d). However, in BOTTOM-TOP and TOP-
BOTTOM scenarios, water movement was not homogenous and that is why the breakthrough
curve for these two cases had a sharp rise compared to other scenarios. This observation was
supported by velocity vector plots [Figure 4(b) and 4(c)].

Design Life

Time needed to reach the design effluent standard for arsenic of 0.1 mg/L varied between 10
– 20 days (TOP – TOP condition was not considered due to overflow issue) [Figure 3(b)].
Hence, these small scale wetlands can be used for 10 – 20 days for treating arsenic laden
wastewater (influent arsenic concentration 1 mg/L). The simulation was carried out at a
constant flow condition of 50 cm2/day. Before breakthrough (10 – 20 days) is observed, a
small scale wetland can treat around 6000 – 12000 cm3 of arsenic laden water. Based on
these data, a field scale wetland which is 1165 times larger than the small scale wetland can
treat 7000 – 12,000 L of water before breakthrough would occur at an effluent concentration
of 0.1 mg/L.
Preliminary Studies for Designing a Wetland for Arsenic Treatment 199

Water Budget

Average inflow and outflow in the field scale wetland was estimated as 35 L/day and 10
L/day respectively. From precipitation and evaporation the net result was loss of water and it
was 0.7 L/day. Transpiration loss of water was around 24 L/day (Figure 5). HYDRUS 2D
output includes cumulative flux, water and solute flux, and pressure head variation with time
and distance. Among these simulated results, outflow (seepage face flux) and pressure head at
the top surface (atmospheric boundary head) have implications on wetland operation and
design [Figure 6(a) and (b)]. Simulations showed that during April to October the soil was
unsaturated whereas in other times there was a water head (~5 cm) built up over the soil
strata. During the same period (April to October) the outflow was also zero.

Transpiration: 24.5 L/d


Precipitation + Evaporation: 7.35 L/d

Outflow 6.5 L/d

Aerated water
122 cm
38 L/d

610 cm

Figure 5. Schematics of water budget for the field scale wetland.

(a) Seepage face flux (b) Atmospheric boundary head

Figure 6. Seepage face flux and atmospheric boundary head profile.


200 Raja Chowdhury, Defne Apul and Daryl Dwyer

Hence, during that period, air movement through the soil strata would increase and would
change the prevalent redox profile inside the wetland. Therefore, if one wants to keep the
wetland saturated all over the year, increasing the inflow during April to October period
would be essential. The water budget estimation using HYDRUS can be a helpful tool for
regulating the flow and thus controlling the redox profile in wetlands.

Conclusion
• Adsorption isotherms for the three soils measured were best modeled by the
Freundlich isotherm.
• Hydraulic conductivity and dispersion coefficient estimated for nusoil were 0.5
cm/day and 0.71 cm2/hr respectively.
• Residence time of arsenic was not much affected by the inlet and outlet position.
However at top inlet and top outlet condition overflow of water was observed.
• Water budget estimated from the simulation showed that at constant head boundary
conditions, in the period between April-October, the wetland would become
unsaturated.
• The volume of water a field scale wetland can treat was found to be around 582 –
1165 L.

References
Bar – Yosef, B.; Chang, A.C.; and Page, A. L (2005). Mass balance modeling processes in
croppland soils. Environ. Geochem. 27:177 – 184.
Buddhawong, S.; Kuschk, P.; and Mattusch, J.; Wiessner, A. and Stottmeister, U. (2005)
Removal of arsenic and zinc using different laboratory model wetland systems. Eng. Life
sci. 5(3): 247 – 252.
Cohen, R.R.H (2006) Use of microbes for cost reduction of metal removal from metals an
mining industry waste stream. Jor. clean. Produc. 14: 1146 – 1157.
Crites, R.W.; Middlebrooks, E. J.; and Reed, S. C (2006) Natural wastewater treatment
systems. Boca Raton, Taylor and Francis.
Dousset, S.; Thevenot, M.; Pot, V.; Simunek, J.; Andreux, F (2007). Evaluating equilibrium
and non – equilibrium transport of bromide and isoproturon in distributed and
undistributed soil columns. Jor. Contam. Hydrol. 94: 261 – 276.
EPA (Environmental Protection Agency) 2002 Arsenic treatment technologies for soil, waste,
and water. Solid waste and emergency response (5102 G) EPA – 542 – R – 02 – 004.
www. Epa.gov/tio.
Haque, N.; Morrison, G.; Cano – Aguilera, I.; Gardea – Torresdey, J. L (2008) Iron –
modified light expanded clay aggregates for the removal of arsenic (V) from
groundwater. Microchem. Jor. 88: 7 – 13.
Hiley, P (2003). Performance of wastewater treatment and nutrient removal wetlands
(reedbeds) in cold temperature climates, pp. 1 – 18 In Advances in ecological sciences
series 11. Constructed Wetlands for wastewater treatment in cold climates, eds Mander,
U. and Jenssen, P. WIT press.
Preliminary Studies for Designing a Wetland for Arsenic Treatment 201

ITRC (2003) Technical and regulatory guidance document for constructed treatment
wetlands. Prepared by The Interstate technology and regulatory council wetland team.
ITRC (Interstate Technology & Regulatory Council). 2005. Characterization,
Design,Construction, and Monitoring of Mitigation Wetlands. WTLND-2. Washington,
D.C.:Interstate Technology & Regulatory Council, Mitigation Wetlands Team. Available
on the Internet at http://www.itrcweb.org.
Interstate Technology & Regulatory Council (ITRC). 2006. Planning and Promoting
Ecological Reuse of Remediated Sites. ECO-2. Washington, D.C.: Ecological Land Reuse
Team, Interstate Technology & Regulatory Council, www.itrcweb.org.
Jong, T.; and Parry, D.L (2003). Removal of sulfate and heavy metals by sulfate reducing
bacteria in short – term bench scale upflow anaerobic packed bed reactor runs. Wat.
Res.37: 3379 – 3389.
Jong, T.; Parry, D. L. (2004) Adsorption of Pb (II), Cu (II), Cd (II), Zn (II), Ni (II), Fe (II), As
(V) on bacterially produced metal sulfides. Jor. Colloid and Interface Sci. 275: 61 – 71.
Liu, W. J.; Zhu, Y. G.; Hu, Y.; Williams, P.N.; Gault, A. G.; Meharg, A. A.; Charnock, J. M.;
and Smith, F. A. (2006) Arsenic sequestration in iron plaque, its accumulation and
speciation in mature rice plants (Oryza Sativa L.) Environ. sci. technol. 40: 5730 – 5736.
Mohan, D.; and Pittman, C. U (2007) Arsenic removal from water/wastewater using
adsorbents – a critical review. Jor. Hazord. Mater. 142: 1 – 53.
Koch, S M.; and Rawlik, P.S (1993) Transpiration and stomatal conductance of two wetland
macrophytes (Cladium jamaicense and Typha Domingensis) in the subtropical
everglades. Amer. Jor. Bota. 80(10): 1146 – 1154.
National Oceanic and Atmospheric Association (NOAA) published atmospheric data Toledo
Ohio.
Ouyang, Y (2005) Phytoextraction: Simulating uptake and transpiration of arsenic in a soil –
plant system. Intern. Jor. Phytorem. 7(1): 3 – 17.
Peake, M.B (1996) Modeling plant uptake of metals in constructed wetlands supported by
experimentally derived uptake rates. Master thesis. Air force Institute of technology.
Rajmohan, N.; Elango, L (2007). Mobility of major ions and nutrients in the unsaturated zone
during paddy cultivation: a field study and solute transport modelling approach. Hydrol.
Proces. 21(20) 2698 – 2712.
Rofkar, J. R.; Dwyer, D. D.; Frantz, J. M (2007) Analysis of arsenic uptake by plant species
selected from growth in northwest Ohio by inductively coupled plasma – optical emission
spectroscopy. Comm. Soil Sci. 38(17 – 18): 2505 – 2517.
Stein, O. R.; Borden – Stewart, D. J.; Hook, P. B.; Jones, W. R (2007) Seasonal influence on
sulfate reduction and zinc sequestration in subsurface treatment wetlands.Wat. res. 41:
3440 – 3448.
Sawyer, C. N.; McCarty, P. L.; and Parkin, G. F (2003) Chemistry for environmental
engineering and science. (5th Ed.) Tata McGraw – Hill Publishing Company Ltd. New
Delhi.
USEPA (1997) Office of solid waste and emergency response. Recent development for in –
situ treatment of metal contaminated soils; U.S. Government printing office: Washington,
DC, EPA – 542 – R – 97 – 004.
USEPA (2000) Manual: Constructed wetlands treatment of municipal wastewaters,
EPA/625/R – 99/010.
202 Raja Chowdhury, Defne Apul and Daryl Dwyer

Verma, P.; George, K. V. Singh, H.V.; Singh, R. N (2007). Modeling cadmium accumulation
in radish, carrot, spinach and cabbage. Appl. Math. Model. 31: 1652 – 1661.
Wohling, T.; Schmitz, G. H (2007) Physically based coupled model for simulating 1D
surface-2D subsurface flow and plant water uptake in irrigation furrows. I: Model
development. Jor. Irrig. Drain. Engrg. ASCE 133 (6): 538 – 547.
Ye, Z. H.; Lin, Z. Q.; Whiting, S. N.; de Souza, M.P.; and Terry, N. (2003) Possible use of
constructed wetland to remove selenocyanate, arsenic, and boron from electric utility
wastewater. Chemosphere 52: 1571 – 1579.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 203-229 © 2008 Nova Science Publishers, Inc.

Chapter 8

GROUNDWATER MANAGEMENT IN THE NORTHERN


ADRIATIC COAST (RAVENNA, ITALY):
NEW STRATEGIES TO PROTECT THE COASTAL
AQUIFER FROM SALTWATER INTRUSION

Beatrice M.S. Giambastiani1,a, Pauline N. Mollema1,b


and Marco Antonellini1,c
1. IGRG-CIRSA (Integrated Geoscience Research Group - Interdepartmental Centre for
Environmental Sciences Research), University of Bologna, Via S. Alberto 163,
48100 Ravenna – Italy

Abstract
The Emilia-Romagna Adriatic coast represents a strategic economic asset for Italy
because of its important environmental, historical and tourist value. An efficient
environmental protection and water resource management in this coastal zone is difficult
because of the many events that during the last century caused the ingression of large volumes
of brackish and saline water in the coastal aquifer.
River mouth enlargement to accommodate for tourist marinas has caused extensive
landward saltwater encroachment along rivers and canals. Tourist establishments along the
coastline have destroyed the original sand dunes continuity and their natural barrier effect to
beach erosion and saltwater intrusion. Strong natural and artificial subsidence, induced by gas
winning and deep groundwater exploitation, caused most of this territory to drop below mean
sea level and have modified the river regime and the normal groundwater flow.
In this situation, a drainage system is necessary to lower the water table level and keep
the land dry, so that trees roots stay above the saturated zone of the aquifer. The drainage
system management, however, has not accounted for the proximity to the sea. The result is an

a
E-mail address: beatrice.giambastiani@unibo.it, +39 0544 937318.
Current work address: UNSW (University of New South Wales), Faculty of Science, School of Biological, Earth
and Environmental Science, 2052, Sydney (NSW). E-mail address: b.giambastiani@unsw.edu.au, +61 2 9385
1208
b
E-mail address: pmollema@gmail.com
c
E-mail address: m.antonellini@unibo.it, +39 0544 937318
204 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

unstable phreatic aquifer that is not able to contrast saltwater intrusion and is not beneficial to
maintaining healthy pine forests and avoiding soil salinization.
Water table and surface isosalinity maps have been created by monthly monitoring data.
These data show watertables below sea level and groundwater salinization; well testing and
geo-electric surveys provided a detailed lithological characterization and the depth of the
brackish-freshwater interface.
Analytical and numerical modeling (WATBAL, SUTRA, MOCDENS3D, and
SEAWAT) were used to calculate the water budget and study how past and present human
activities have affected the saltwater intrusion process and how the predicted future sea level
rise will further degrade the aquifer.
The feasibility and effects of some alternative actions such as controlled drainage,
freshwater storage in ponds along the coast, artificial recharge, river bottom sills, optimization
of the hydraulic infrastructure and coastal dune restoration have been examined.
In addition, the interactions between surface and ground waters as a connected resource
need to be further explored in order to find innovative strategies for coastal aquifer
management.
Finally, we stress how an integrated approach is necessary; history has shown that a
disaggregated water management, the present fragmentation of water authorities and the lack
of communications among different institutions and the stakeholders are the major threats to
the coastal environment and its water resources.

Introduction
The Adriatic coast of Emilia-Romagna in Northern Italy represents a strategic economic asset
for Italy, because of its important environmental, historical and tourist value. The present
hydraulic infrastructure is a dense network of rivers and canals, wetlands, lagoons, a complex
layout of urban areas, tourist establishments, industrial infrastructures and agricultural
drainage systems, all coexisting on a small area (Fig. 1).
Originally, the coast of Emilia-Romagna formed the edge of alluvial plain fed by
sediments and waters from rivers coming from both the Apennines and Alpine mountain
chains of which the river Po is the largest and most famous contributor. Natural changes in
the course of these rivers and of relative sea level changes brought alterations to the coastline
and the deposits along it, including the layout of river mouths, lagoons, dunes and beaches.
Thus, some of the sediments that now form further inland the bottom of the aquifer were
originally formed as deposits at sea in front of the ancient river mouths, and as a consequence
may still contain some old connate seawater.
In these coastal areas many man-made changes were made predating the Romans to make
the land better suit the needs of the people. For example, a large part of the coastal plains used
to be covered with swamps and various techniques of land reclamation have made many of
those swamps dry. The strong natural subsidence typical of a natural delta was enhanced even
more by deep groundwater exploitation and gas winning in the last century.
Table 1 lists these and some other of the most important natural and man-made changes
that happened in the last two thousand years with reference to Figure 1 for the location.
Figure 1. Location of the study areas along the Emilia–Romagna Adriatic coast: 1) Comacchio Lagoon and Vene Ancona – Bellocchio, 2) Marina Romea,
Piallassa Baiona lagoon, San Vitale pine forest, Punte Alberete and Valle Mandriole wetlands (zoom on the left); 3) Classe pine forest and Ortazzo-Ortazzino
wetlands. Note: the complex system of rivers, canals, lagoons, urban area and the large area below sea-level. The triangles indicate the extend of saltwater
encroachment along the main rivers: triangle (1) along Po di Goro, 25 km from the river mouth, (2) along Po di Volano, 9 km from the river mouth and (3)
along the Lamone river, 3.8 km from river mouth (modified from Antonellini et al., 2007).
Table 1. Overview of some of the most important natural and man made events affecting the hydrology of the unconfined coastal
aquifer, Emilia-Romagna, Italy .

EVENT BEGINNING END NOTES AND REFERENCES


Planting first pine forests Augustus period ? By Romans for wood for shipbuilding
Cervia salt mine formation 6th century AD 9th century AD Etruscan salt mine still operating today
Planting pine forests (such as historical 10th-15th centuries AD 19th century AD (when Planted by monastic communities on fossil dunes
Cervia, Classe and San Vitale pine their destruction started)
forests still existing)
Coastal dune formation Fossil dunes on old
coastlines: 10th-18th
centuries AD
Dunes closest to shore:19th
century AD
Piallassa Baiona lagoon formation 17th century AD 19th century AD Formed by an increase in sediment supply (17th-18th
centuries AD) that formed a new peninsula closing off
part of the sea and by artificial interventions related to
construction of the port of Ravenna (Bondesan, 1990)
Land reclamation, drainage system 19th - 20th centuries AD 1960-1970 Men broke the dikes and let the river deposit its
(In 1839 Lamone flooding sediments on the land to counteract the natural
marked the beginning of subsidence
land reclamation for the The mechanical drainage continues in order to keep
study area) agricultural areas dry.
Works along the rivers Reno and 1830 1968 These works caused strong reduction in sediment
Lamone transport and as a result changing coastlines between the
months of these rivers and in area Vene Ancona-
Bellocchio (Bondesan et al., 1978, 1995(b); Simeoni et
al., 2002)
Water extraction directly along shore 1930 in Rimini On going Many illegal wells still exist although most tourist
and in dune for tourist establishments After 1950 in Marina establishments should be connected to regional drinking
Romea water system.
Table 1. Continued

EVENT BEGINNING END NOTES AND REFERENCES


Planting coastal pine forests 1933 Planted by Senator Luigi Rava along the current
coastline in order to protect inlands from sea flooding
and marine aerosol
Anthropogenic subsidence 1950 1970-1980 By gas winning and deep groundwater exploitation
during the industrial development of Ravenna
From 1985 the subsidence decreased and today is 3-4
mm/year (higher along coast up to 0.8-1.6 cm/year).
(Teatini et al., 2006; ARPA, 2003; Preti, 2000;
Gambolati and Teatini, 1998)
Dune destruction 1954 Still going on Started with the development of tourism.
Construction of dikes around 1960 This is part of the land reclamation and separates the
Comacchio lagoon saltwater lagoons of Comacchio from the agricultural
lands - before all swamps.
River mouth construction of Lamone 1960s A river mouth was already constructed in 1800, later
closed.
During the 1980s-1990s other interventions created the
current situation
Construction of piers at Porto Corsini 1968
Beach Erosion 1970-1980 Still continuing
Declared RAMSAR site Punte Alberete 1976 Residue of the former wetlands presented before the land
and Valle Mandriole reclamation (in 1960s)
Beach nourishment 1980 From 1980-1990 onwards
Every 4-5 years – most recently 2007
Breaking up of peninsula that formed 1981 This event changed the whole coastal morphology
the end of the Reno river mouth moving the river mouth 2 km to the south (Idroser, 1996)
Creation of fresh-brackish water bodies 1990s Constructed to create a buffer zone between the lagoon
inside the Piallassa Baiona lagoon and the pine forests (San Vitale)
208 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Until very recently (2000) the man-made changes were performed typically with one
particular scope in mind: for example drainage to create agricultural land, or drainage to
preserve forests along the coast from flooding (Fig. 2a), or extending a river to sea to prevent
flooding more land inward, preserve saltwater lagoons and ponds for fisheries, exploiting
drinking water by pumping.

a)

b)

Figure 2. (a) Drainage canal inside the coastal pine forest (Marina Romea, Ravenna); (b) example of
coastal erosion and sea flooding near the coastal wetlands of Ancona-Bellocchio.
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 209

Now it is common knowledge that any such single action will have major consequences
for the hydrology of a large area, but at the time this was not considered and so many
unwanted side effects are now causing problems to the area. For example, subsidence coupled
with river canalization and rectification caused rivers to become ‘hanging’ above the
surrounding land, strongly reducing sediment supply to the sea and consequently causing
coastal erosion along most of the Emilia-Romagna shoreline (Fig. 2b). The subsidence also
dropped most of this territory below mean sea level and has modified the river regime and the
normal groundwater flow. In this situation, a drainage system became necessary to lower the
water table and keep the land dry.
River mouth enlargement to accommodate for tourist marinas has caused extensive
landward saltwater encroachment along rivers and canals. Tourist establishments along the
coastline have destroyed the original sand dunes continuity and their natural barrier effect
(Fig. 3).
All these man-made changes (Table 1) reduced in one way or another freshwater supply
in the coastal aquifer. The result is an unstable phreatic aquifer that is not able to contrast
saltwater intrusion. Unluckily, it takes a long time (up to thousand years sometimes e.g. Oude
Essink, 2004) for any aquifer to regain a stable brackish-freshwater interface after a change
has occurred. Therefore our coastal aquifer is still responding to changes that occurred several
hundred years ago or later. This is important to realize especially when modeling studies are
performed. For this very reason we thought it important to understand and document the
hydraulic and hydrologic history of this area.
Saltwater intrusion is not the only hydrologic problem facing the area: droughts, flooding
during events of extreme precipitation, contamination of surface and underground water
resources, and over-exploitation by agriculture and tourism also need to be considered. These
problems will only worsen as current climatic changes continue (Giambastiani et al., 2007,
Antonellini et al., 2008)
Moreover another complicated issue is the fragmented and complicated Italian
bureaucracy of water management. This is historically grown along with the ad-hoc changes
in the hydraulic structure. The so-called ‘Consortium of land reclamation’ (Consortium
Bonifica) for example was called to design and execute the land reclamation starting in the
1950s. To this day, they manage the drainage of the area as well as the irrigation of the
farmland. They make decisions about the quantities of freshwater, but are an entity totally
disconnected with other political entities such as the established water shed authorities, the
Regional water entities, the Province or the communities.
This chapter will first provide a brief summary of the hydrogeological characterization of
the aquifer, its state of salinization and causes of saltwater intrusion based on previous
monitoring and modeling works (Giambastiani et al., 2007, Antonellini et al., 2008, Ulazzi et
al., 2008) and work in progress (Mollema et al., 2008). Possible remediation methods that
could counteract saltwater intrusion will be discussed as well as the challenges that scientists
and water management authorities need to face in order to change a historically grown way of
doing things, into sustainable water management that serves many scopes at once over a large
territory.
210 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

a)

b)

Figure 3. (a) Example of dune destruction along the Marina Romea coastline (Note: the tourist
establishment built on the dune); (b) scheme of the static natural barrier effect of coastal dunes: where
the dune system is not continuous saltwater intrusion easily occurs because the freshwater head is not
able to contrast the saltwater upconing.

Aquifer Lithostratigraphy

The Holocene geomorphic evolution of the Romagna coastal plain has been controlled by
continental (Wurmian) and marine deposition (post Wurmian transgression) in the coastal
environment of the Po River Plain. Rizzini (1974), Amorosi et al. (1999, 2002) and Bondesan
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 211

et al., (1990, 1995(a)) have described the late Quaternary depositional history of this coastal
plain.
The deposits making up the phreatic aquifer are typical of coastal and delta areas with
sandy and silty unit intercalated (Marchesini et al., 2000) and sand dunes characterized by
north-south orientation, which today are visible up to 30 km inland. Aquifer’s thickness
varies from 8 to 30 meters (Antonellini et al., 2008). Generally, the lithologic reconstruction
of the study area shows a dominant sand composition with medium and fine grained sand
units characterizing the most part of the aquifer’s stratigraphy.
Slug tests and tidal well tests have been carried out in the most significant points of the
study area (2004-2007) in order to define the main petrophysical properties and evaluate full
scale aquifer hydraulic conductivity. We used the Bouwer and Rice slug test method
(Bouwer, 1976; 1989) that is suitable for screened wells in the unconfined coastal aquifer.
The results of the slug tests have been integrated with the lithological/stratigraphic
information and compared with tidal well test values of hydraulic conductivity obtained using
the method outlined by Fetter (2001) and Jacob (1950). Hydraulic conductivity estimates give
an average of 20 m/day (Giambastiani et al., 2007). This high hydraulic conductivity values
are in according with the stratigraphic records which show a dominant sand composition with
high porosity (0.15-0.35%) (RER and ENI-AGIP, 1998).

Water Budget

In order to estimate the water budget it is necessary to know all inflow and outflow
components in the catchment. Unfortunately due to the lack of accurate drainage and
irrigation information many data can only be deduced. Therefore the approach has been to
estimate local water budget taking into account only precipitations, evapotranspiration,
runoff, infiltration and pedologic features.
An end-of-the month water balance model called Watbal (Starr, 1999) was used to
calculate the components of the hydrologic water budget for several sites in the study area
(Antonellini et al., 2008; Mollema et al., 2003). This model is based on an energy balance
approach to estimate potential and actual evapotranspiration, soil water flux, soil moisture
content and deficit. The estimate of these parameters depends on several factors such as solar
irradiation, air humidity, wind speed and direction, soil composition and type of vegetation.
The average precipitation is 0.6 m/year and the mean temperature is 13.4°C and generally
the highest temperatures occur in the summer which is also the period with the least
precipitation. As a result the water budget calculations show a surplus only during the first
three months of the year when some water is able to seep down from the soil into the aquifer
below. In the other months of the year, nearly all precipitation is used by plants or evaporates
directly from the soil (Fig. 4)
Since the water to recharge the aquifer is available only a few months per year, the
aquifer recharge is small (14-17 mm/year) and it is insufficient to create a significant
freshwater head.
Moreover Ranjan et al. (2006) prove that pine forests along coastline zone can affected
negatively the water quality in the phreatic aquifer. In fact dense vegetation results in strong
evapotranspiration and decreases the recharge of the aquifer; the pumping made by tree roots
encourages saltwater intrusion from the sea and from the bottom of the aquifer also.
212 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Figure 4. Example of the water budget calculated by WATBAL for the case site of San Vitale pine
forest: values of monthly precipitation (P), potential evapotranspiration (PET), actual
evapotranspiration (AET) and surplus.

Saltwater Intrusion
Current Situation

In order to characterize the aquifer and its current situation, water table and isosalinity maps
have been created by monthly monitoring data from the main areas along the Emilia-
Romagna coast: Marina di Ravenna, Marina Romea, Ancona-Bellocchio, Classe and Cervia.
All data have been used to create a vulnerability map (Fig. 5) where the extension of confined
and unconfined aquifer is shown as well as the depth of brackish-freshwater interface.
For the most part of the case studies water table depth is, generally, a few centimeters
below mean sea level for most of the year. The water table is above sea level only along the
embankments of main rivers such as the Lamone, Reno and along fossil dune belts. Where
the topography is below sea level, the area becomes a polder. Furthermore, due to the
drainage from land reclamation, the hydraulic gradients are directed landward and not
seaward as one would expect in a coastal area (Giambastiani et al., 2007, Antonellini et al.,
2008).
The results of the water salinity monitoring show an extended aquifer salinization until to
the top of the water table in some cases with values similar to those of brackish water (8-15
g/l). The freshwater in the coastal aquifer consists of low salinity water lenses floating on
saltwater and the biggest are below the Classe pine forest, the northern S.Vitale pine forest,
and below the remnants of the coastal dunes (Fig. 5). In the remaining coastal aquifer,
brackish-freshwater interface is 3-4 m deep with high values of salinity in proximity to the
brackish lagoon (Piallassa Baiona, Vene Ancona-Bellocchio, Ortazzo-Ortazzino) due to the
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 213

infiltration from the embankments and saltwater intrusion from the bottom. These wetlands
and inland lagoons can become completely saline or hypersaline water bodies (35-40 g/l)
during the summer period due to the intense evaporation rates.

Figure 5. Vulnerability map of the confined and unconfined aquifer inside the Holocenic and
Pleistocenic sand units. Note: arrows indicate the extend of saltwater encroachment along the main
rivers (modified from Antonellini et al., 2008)

High groundwater salinity values are also recorded close to the pine forests and drainage-
pumping machines, probably as a consequence of the saltwater upconing caused by the trees’
evapotranspiration (Ranjan et al., 2006) and the land reclamation drainage.
Moreover the saltwater encroachment affects the majority of rivers and canals connected
to the sea and generally high value of salinity has been measured inland, until the first weirs
or dams (Fig. 5).
Where sand dunes have been destroyed the coastal protection is reduced and winter
storms can cause seawater inundation and can open a connection between the sea and the
backshore marshes and create new brackish-salty lagoon such has happened in Vene Ancona-
Bellocchio in 2005 (Fig. 6).
214 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Figure 6. Example of seawater inundation caused by a winter storm and formation of washover fan
(Vene Ancona-Bellocchio in 2005)

Geo-electric resistivity surveys (VES and tomography) in the most significant points of
the study area have been carried out in order to determine the location and shape of the
brackish-freshwater interface.
The survey results suggest that the brackish–freshwater interface, defined as the depth
where resistivity is about 8-10 ohm*m (Burger, 1992; Loke, 2001), is close to the surface; its
average depth is 5-6 m, but it deepens to 11 m in depth where high-infiltration recharge areas
are present, like in the fossil dunes of the pine forests (Sabia et al., 2005). Along the coastline
the brackish–freshwater interface does not reach the bottom-confining layer and does not
prevent saltwater intrusion into the aquifer.
In the main tourist areas such as Marina di Ravenna o Marina Romea, the recharge
pattern is affected by the presence of urban and industrial areas. Generally the freshwater
needs along the coast are supplied by water from the foothills of the Apennines (Emilia-
Romagna’s aqueduct) and from the main rivers such as Lamone and Reno rivers. Unlicensed
wells are acknowledged to exist but their pumping rates are not known. It is believed that they
are most active during the summer to meet the water needs of the tourist industry and the
irrigation.

Analytical and Numerical Simulations

Analytical and numerical modeling of this area have been started firstly to better understand
the process of saltwater intrusion in the area and to refine our knowledge of the water budget,
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 215

the aquifer recharge and study the influence of topography and lithology on the brackish-
freshwater interface depth.
The analytical Dupuit-Ghyben-Herzberg relationship (Dupuit, 1863; Bear et al., 1999;
Fetter, 2001) was used to model the position of the brackish-freshwater interface for the area
of Punte Alberete e San Vitale. We have considered a section from the coast to inland areas
and the estimated recharge along the section was calculated according to measured hydraulic
conductivity, the different land uses, runoff and recharge. The permeability was determined
based on the stratigraphic information given by well logs and the resistivity surveys. The
results of this analytical modeling show that the location of the brackish-freshwater interface
is compatible with the boundary surface obtained by the resistivity measurements, with an
average depth of 5-7 m below mean sea level. The maximum interface (10-13m) depth is
registered in proximity of the fossil dunes and the minimum (2-4 m) closed to the brackish
Piallassa Baiona lagoon.
Various runs of the model using different permeability and recharge values show that the
depth of the interface is controlled much more by the permeability of the aquifer than by the
amount of annual recharge.
Variable density groundwater flow modeling studies are carried out to study how the past
and present human activities have affected the saltwater intrusion process in the phreatic
aquifer and how the predicted future sea level rise will affect the salinization process.
Two-dimensional models that aimed to quantify the relative contribution to saltwater
intrusion by various boundary conditions such as channels, rivers, lagoons or topography,
showed that it is very important to simulate the hydraulic and hydrologic history of the area.
So many natural and man-made changes have altered the hydrology relatively recently (see
Table 1), that most probably the flow regime is still responding to those changes and the salt
distribution is also still changing as a consequence. The simulations carried out with
MOCDENS3D (Oude Essink, 1998; 1999; 2001; Giambastiani et. al, 2007) to quantify
hydraulic head, salinity distribution, seepage and salt load fluxes to the surface water system
in the coastal aquifer of Ravenna in a cross section 8 km long, show that over the last century
artificial subsidence and heavy drainage started the salinization process in the study area. The
local climatic conditions and the lack of a continuous coastal dune system favor salt wedge
intrusion.
Considering the model outputs, the salt load will affect the entire coastal hydrogeological
system and groundwater flow during the coming decades when sea levels will rise; the mixing
zone between fresh and saline groundwater will be shifted 800 m farther inland. Since the soil
becomes more saline, farmland degradation and problems for the pine forests would also
occur.
A 3 dimensional model was built for an area of 400 by 500 m with SEAWAT (Guo
and Langevin, 2002) to study the effect of dune destruction on saltwater intrusion
(Mollema et al., 2008). The conclusions are that for the area in consideration near Marina
Romea the overall topography with a continuous second dune belt parallel to the first
interrupted dune belt and the recharge are sufficient to sustain freshwater underneath the
dunes. The discontinuity of the dunes closest to shore or a drainage channel at the back of
the dunes (Fig. 2a) cannot explain the high salinity of the groundwater only 200 m more
inland and so there must be other factors contributing to the salinization: the
evapotranspiration of the trees or seepage from the river Lamone to the north or
mechanical drainage.
216 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

So far the modeling studies have helped to understand better the hydrology of the area
and the factors that enhance saltwater intrusion, but more modeling studies both on a large
scale (20 km) and on smaller scale (100’s of meters) are under way to refine our
knowledge. Large scale models for example are needed to study the effect of the likely
presence of connate salty groundwater at the bottom of the aquifer; until now this effect has
been largely ignored. Small scale models will have to quantify the interaction between
surface waters and aquifer, for example between rivers and aquifers or drainage channels
and aquifers.
Figure 7 shows an example of groundwater modeling studies currently under way using
SEAWAT code. Although the model still needs to be calibrated and compared with
monitoring data, especially at depth, it illustrates the complexity of the system and processes
that need to be further investigated. Figure 7c shows the salt concentration distribution in the
aquifer after 275 years of simulation confirming the presence of freshwater under dunes and
hypersaline water underneath the Piallassa Baiona lagoon. Moreover the model output shows
the upconing of connate salty groundwater from the bottom up to the surface of the aquifer.

Figure 7. Groundwater model of the study area using SEAWAT code; the groundwater system consists of
a 2D profile perpendicular to the coast, 20 km long and 30 m deep, and includes the dunes, the brackish-
salty Piallassa Baiona lagoon, San Vitale pine forest and agricultural lands at the back: a) site geometry of
the aquifer, hydraulic conductivity and thickness of layers as used in the numerical model; b) initial
distribution of salinity concentration, recharge and evaporation rates and location of the observation wells;
c) final salt concentration distribution (g/l) in the aquifer after 275 years of simulation.
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 217

It will be difficult to create models that will exactly reflect or predict measured salinity
values in a particular place and time. This discrepancy of measured and modeled data already
observed (Giambastiani et al., 2007) is due to the lack of sufficient monitoring data,
especially at depth, and due to strong variations in salinity that are caused by weather events
(draught or rain) and that do not show up in the models. The complexity of the area with
many hydraulic structures close to one another further worsens the situation.

Causes and Effects

Our studies indicate that the natural and anthropogenic processes involved in saltwater
intrusion along the Northern Adriatic coast are:

Coastal erosion. Due to the scarce sediment supply (ARPA 2002), most of the Adriatic
coastline is under erosion. Many coastal defenses have been used to preserve the beaches
(dikes, breakwaters and artificial nourishments) but the erosion is not halted. Coastal erosion
brings the sea inland, diminishes the areas of recharge, and the volume of the aquifer
contributing to the increase in groundwater salinization.

Drainage. Land reclamation drainage to allow for agriculture causes a drop in hydraulic
head and aquifer’s freshwater recharge allowing for the upconing of the brackish-freshwater
(Fetter, 2001).

Well pumping. Some coastal zones not yet connected to the aqueduct continue using
water coming from wells in the phreatic aquifer for sanitary and domestic use. Lowering the
piezometric level in proximity to the wells allows saltwater upconing into the aquifer from
underlying higher density saline waters (Bear et al., 1999).

Coastal dune destruction. Dunes are the first line of defense against seawater ingression
especially in an area where most of the territory is below mean sea level. In addition, thanks
to their topographic elevation and good infiltration capacity, dunes provide freshwater
recharge and act as a hydrostatic barrier against the saltwater intrusion according to the
Ghyben-Herzberg principle (Fetter, 2001). The lateral continuity of the dunes system is also
important for creating a continuous freshwater lens along the coast. Unfortunately, the dunes
have been almost completely destroyed by tourist development, building of roads, sport
resorts and parking lots close to the beach.

Encroachment of saltwater along rivers and canals open to sea. Saltwater intrusion
occurs in the groundwater as well as along surface water bodies (Schijf and Schonfeld, 1953).
Due to the lack of proper bathymetric gradients (land subsidence reversed the gradients in
proximity to sea), the rivers in the study area have low discharge velocities. Their mouths
have been deepened and widened to permit the construction of tourist marinas. Furthermore,
rivers provide most of the irrigation water for agriculture, so that the natural river discharge is
reduced almost to zero during the irrigation season (May-September). In this situation,
saltwater encroachment is facilitated; once the salt wedge has penetrated far enough inland, it
can pollute the phreatic aquifers in contact with the river’s bottom (Schijf and Schonfeld,
1953; Armi, 1986; Baines, 1995)
218 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Land subsidence. Surface elevation is important in controlling the depth of the saltwater-
freshwater interface (Fetter, 2001; Hubbert, 1940). In the last century, natural (Selli and
Ciabatti, 1977; Pieri and Groppi, 1981; Carbognin et al., 1995; Brambati et al., 2003;
Carminati et al., 2005) and anthropogenic subsidence (Preti, 2000; Teatini et al., 2006)
caused important topographic variations of the area with fast subsidence rates. In some areas
the total subsidence in the period 1950-1990 has been up to 0.9-1.0 m (ARPA 2002).
Measures were taken to limit groundwater and gas extraction in the late 1980s and the rate of
subsidence has since decreased. Land subsidence has dropped most of the territory below
mean sea level modifying the river and normal groundwater flow regimes. This process has
aggravated the problem of the small hydraulic gradients towards sea.

Insufficient aquifer recharge. The aquifer recharge is limited by urbanization and river
straightening/channeling. River rectification, in fact, in order to gain more land for
agriculture, has reduced the recharge from the rivers to the underlying phreatic aquifer.

Hydraulic Infrastructure. The local hydraulic infrastructure has been built with the
major objective of defending the territory from flooding or for irrigation and land reclamation
purposes. Its management therefore has never been geared to ward, also prevent saltwater
intrusion. This fact, coupled with a poor maintenance and efficiency of the hydraulic
infrastructures, contributes to a disaggregated management of water resources in the coastal
zone.

Integrated Coastal and Water Resource Management


An efficient water resource management for the Emilia-Romagna Adriatic coast is a complex
endeavor, because during the last century, many events, as shown in Table 1, have irreparably
affected the natural equilibrium.
History has shown that the prior disaggregated management, the present fragmentation of
water authorities and the scarce communication among multiple stakeholders can be a threat
to the coastal environment and its water resources.
During our research, we evaluated and proposed the feasibility of some alternative water
management actions. In particular, we evaluated some actions already taken and other to be
taken in the near future in order to prevent saltwater intrusion. These measures include
controlled drainage, freshwater storage in basins along the coast, artificial recharge, coastal
dune restoration and renewal of hydraulic infrastructures.

Controlled on-Demand Drainage

The estimate of the land reclamation drainage is not easy due to the lack of accurate
information and long time series. Considering currently data (1999-2004), we have correlated
annual pumping water (Q), basins extension, rainfall (P), and effective evapotranspiration
(AET) (Starr, 1999) to evaluate if pumping could be better managed during the year with
respect to the current practices. It results that almost every year there is a deficit ranging from
20 to 100 mm of water due to the combined effect of AET and over-pumping from the land
reclamation infrastructure. Generally, the amount of pumping water added to the
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 219

evapotranspiration is bigger than the precipitations and only in a few cases small quantity of
freshwater (1-3 cm/year) is available for the infiltration and aquifer recharge. Over-pumping
from this infrastructure could be avoided by storing the water in surface basin currently
abandoned and releasing it on-demand for irrigation. The drainage has a positive effect on
keeping the farmland and the forests along the coast dry but it has a negative feedback effect
on aquifer salinization reducing the hydraulic head necessary to contrast saltwater intrusion.

Freshwater Storage at the Surface

In the San Vitale pine forest, an attempt at freshwater storage in basins along the coast was
adopted in 1996, in order to create a buffer zone between the brackish-salty lagoon and the
forest. We have evaluated this artificial intervention to see its effect on the depth of the
brackish-freshwater interface. Figure 8a shows the location of two adjacent profiles in the San
Vitale pine forest; the first profile (1) starts at the shore of the water body that in 1996 was
filled with freshwater, the second profile (2) starts at the dike separating the freshwater body

Figure 8. a) Location of the profiles (1 and 2) used to evaluate the effects of the freshwater surface
bodies along the eastern side of the San Vitale pine forest. The pine forest is in black, the saltwater
lagoon in grey and the freshwater ponds are in white; b) profile 1 in which the brackish-freshwater
interface is calculated assuming there is no freshwater pond. The calculated interface is shallower than
the measured interface; c) profile 2 is longer than profile 1 and includes the part across the freshwater
pond until the edge with the rest of the saltwater lagoon. The calculated depth of interface matches the
measured interface suggesting the freshwater pond was effective.
220 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

from the seawater. We computed the depth of the brackish-freshwater interface along these
two profiles using the water budget data, the aquifer properties measured from well tests
(Antonellini et al., 2008), and the Ghyben-Herzberg-Dupuit relationship (Fetter 2001). The
depth of this interface in the two models was compared with the depth of the interface
measured in two piezometers along the profiles (white squares in Fig.8b and 8c). Comparison
of figure 8b and 8c shows that the interface in the case of no freshwater pond would be much
shallower (Fig. 8b) than it is now in the presence of freshwater pond. The data we measured
today match well the case shown in figure 8c, therefore we can conclude that the freshwater
ponds built in 1996 have contributed to slowing down seawater intrusion.
The freshwater storage in ponds along the coast can be useful only if the hydraulic head
is always kept higher than sea level. In this case it is possible to increase the freshwater
recharge and create a natural local barrier against the saltwater intrusion. Moreover, according
to the local climatic conditions, the high evaporation from these ponds and the salt
concentration process during the summer period needs to be considered.

Artificial Injection Recharge System

An injection recharge system (Giambastiani et al., 2006) along the cost or close to brackish
lagoon and/or wetlands would be feasible but at a high cost, because the elevated
permeability in this coastal area requires large amount of injection freshwater, which actually
is not available, to provide a significant increase in hydraulic head and to make the brackish-
freshwater interface reach the bottom of the aquifer creating, in this case, an efficient barrier
system to contrast the saltwater intrusion (Fig. 9).
Given the cost of water at a regional level, the cost of such a solution is expensive; a
more economic recharge system would be possible only where the hydraulic conductivity
values are small.
Furthermore, the effect of a water table rise may be negative for lowlands vegetation and
the roots of the pine trees.

a)

Figure 9. Continued on next page.


Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 221

b)

Figure 9. (a) Artificial recharge system considered along the coast or close to brackish lagoon and or
wetlands: wells 200 m spaced (d) and 25-30 m deep (h);(b) K indicates hydraulic conductivity values;
Q are correspondent injection freshwater rates needed to create a hydraulic head (black line) above sea
level between the wells and make the brackish-freshwater interface (grey line) reaching the basement
aquifer. The pumping rate is very sensitive to K and inter-well spacing. Where K is large we need also a
large Q or a short inter-well spacing.

Coastal Dunes

As the results of the modeling studies as well as monitoring show, the remaining dunes still
contain freshwater. The preservation of the existing topography that is above sea level,
including the dunes along the coast is crucial for the survival of the freshwater aquifer
(Minchio et al., 2007). Not only are the dunes directly along the waterline important but also
older dune belts or positive topography further away from the waterfront are important in
contrasting saltwater intrusion (Mollema et al., 2008). Although subsidence rates have
decreased largely since the 1980s as ground water and gas exploration was halted, subsidence
rates along the coast can still amount up to 0.8-1.6 cm/yr (ARPA 2003). This means that a lot
of territory may disappear below sea level within the next 50 – 100 years. Because of these
high subsidence rates, accurate topographic maps are not available for the whole area and
perhaps a first step is to accurately document which of the areas are above and below sea
level now and in the future. LIDAR data could be helpful if GPS or other monitoring data are
unavailable (Mollema et al., 2008). Management strategies of the dunes and areas close to
shore that typically include the pine forests should focus on how to stabilize the topography
and, at the same time, how to increase the recharge of water into the underlying aquifer.
Dunes typically are an area of very high permeability and high recharge but along the
Adriatic coast the pine forests may use a large part of the precipitation leaving little or no
water available for the aquifer recharge. However some vegetation is needed to stabilize the
new and old dunes and many factors such as precipitation, temperature, evaporation and wind
magnitudes (Tsoar 2005) need to be taken into account to determine the best vegetation cover
that does not use too much water yet stabilizes the topography.
222 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

The restoration of dunes that were leveled to make place for bathing may further help to
increase the freshwater supply. As beach erosion is already very strong, many times beach
nourishment projects were carried out with sand coming from the bottom of the Adriatic Sea
(Table 1). Perhaps these nourishments should be combined with dune restoration as well.
Emilia-Romagna dunes are all in highly developed tourist area with huge economic
interests. One of the most difficult tasks of the regional authorities will be to prevent further
dune destruction to create space for more tourist establishments. Also, at the end of the tourist
season existing establishments often bulldoze the sand from the dunes towards the front of the
establishments as protection against winter storms. This practice further contributes to the
breakdown of the dunes and freshwater aquifer destruction and should be prevented. The
public and authorities will need to be educated that it is in the interest of everybody to
preserve the coast with dunes and freshwater underneath.

Hydraulic Infrastructures

One of the most effective solutions, at the present time, is the maintenance and optimization
of the existing hydraulic infrastructure and the building of weirs and artificial sills along
rivers and canals open to sea in order to allow discharge to the sea in the winter and avoid
saltwater intrusion in the aquifer and the landward saltwater encroachment in summer and/or
during high-tide periods.

Figure 10. Measure to avoid the saltwater seepage along the river banks: (a) building of a second weir
closer to the river mouth and (b) installation of impermeable anti-seepage walls, down to the aquifer
basement.
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 223

It is needed to move the weirs seaward from their current position and to make the
remaining stretches of the river more impermeable by constructing clay barriers in the dikes
that can reduce the hydraulic conductivity of the river’s banks. Currently impermeable anti-
seepage walls (formed by water, bentonite and concrete) have been used along the Lamone
river banks to reduce the saltwater intrusion from the riverbed into the surrounding farmland
(RER, 2006) (Fig. 10).
Unfortunately the impermeable walls in the river banks have not been effective, because
the impermeable layers are 8 m deep from the surface and cannot prevent saltwater upconing
from the river bottom. Impermeable walls solutions are expensive because they should reach
the aquifer basement in order to stop, or at least reduce, the saltwater infiltration.
Apart from these hydraulic works, it is necessary to maintain the efficiency of the already
existing infrastructures in order to prevent the saltwater encroachment along the river.

Land Uses

Finally, an integrated coastal management needs to consider the optimization of land use,
taking into account the future climatic changes and the effects of the rainfall distribution
changes on the evapotranspiration and hydrologic balance. A correct land use based on soil
composition, less water-intensive crops and proper water conservation could reduce water
needs in specific areas.

Discussion and Conclusion


As recent research (IPCC, 2007) predicts, a reduction of freshwater resources and an increase
of drought frequency and flash floods will occur in the coming 100 years. The European
Water Framework Directive (WFD) require water authorities to take into account the multi-
functionally and diversity of water resources in any integrated water management plans (EC,
2000).
The saltwater intrusion problem along the southern Po plain coast has been ignored for a
long time as the freshwater was not considered an important and valuable resource but a
“given” for the stakeholders in the coastal economy. In general, the current use of freshwater
for civil and agriculture use as well as for the maintenance of natural habitats is unsustainable.
Until today, the regional authorities have not considered the coastal phreatic aquifer as a
strategic water reservoir for drinkable water supply and it has received fewer attentions than
the other groundwater in the deeper confined aquifers (Di Dio 1999). A radical change in
attitude on all fronts – political, economic and scientific - must be adopted if further
salinization of the coastal aquifers is to be prevented, especially considering future climatic
changes. The climate change will bring variations in the hydrologic cycle and increase the
pressure on the coastal aquifer. The future sea level rise and higher average temperatures will
enhance the loss of freshwater due to increased evapotranspiration and that would have a
major effect on the water supply for irrigation and drinking water. It will also alter many of
the habitats and ecological communities (coastal pine forest and wetlands) posing a danger
for existing plants and animals.
224 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

As the modeling simulations show, the future sea level rise will accelerate the increase in
salt load and seepage affecting the entire unconfined coastal aquifer with a very shallow and
inland brackish-freshwater interface. As the soil becomes more saline, farmland degradation
and vegetation growth problems would also occur.
An integrated management in our study area needs to consider:

• Lack of water during the year compared with the variable demand during the seasons
for tourism, agriculture and industry;
• Management of drainage and irrigation in the farmland area;
• Freshwater-saltwater interface management to preserve freshwater wetlands
ecosystem and to avoid soil salinization;
• Coastal dune conservation and restoration;
• Interactions between groundwater and surface resources traditionally treated as
separated resources;
• Sustainability of the different water management strategies in the local economy;
• Role of each stakeholder within the EU and local regulations framework.

Remediation and preventative actions should be taken in time and planned on a long time
basis and large space scale. Integrated coastal management should consider the aquifer, coast
and dunes as a single and complete system and should take into account the possible feedback
effects of each action. Generally, short term strategies are often adopted to solve emergency
situations but long term strategies require a complete knowledge of the physical and
ecological process so as to be flexible enough to adapt to the changing climate scenarios
(Antonellini et al., 2008). In addition, the interactions between surface and ground waters as a
connected resource need to be more explored in order to find innovative strategies to protect
the coastal aquifer and avoid their current management as separate resources. This behavior
typically results in the separate estimation of the sustainable yield of surface and groundwater
systems when, in fact, their yields are interdependent (IAH, 2004). In fact surface-ground
interaction can have significant implications for both water quantity and quality; water flow
regimes, water security, aquatic ecology, salinity and nutrient loading can all be affected by
the flow of water between surface water and underlying aquifer, especially where this last one
is polluted by saltwater. In this case, there needs to be improved modeling of surface and
ground waters interconnectivity, increase density of monitoring boreholes and rainfall
gauging stations in order to improve data collection.
Further investigations can be applied and other issues need to be considered in our study
area in order to advance the scientific understanding of the aquifer and help policy makers to
implement more effective management of the scarce freshwater resources in the area:

• Water quality monitoring of irrigation canals;


• Groundwater monitoring network;
• Geophysical surveying for mapping the recharge and discharge zones along the
rivers as well as for mapping the brackish-freshwater interface in the unconfined
aquifer;
• Combining surface and ground water modeling with groundwater chemistry to better
understand the migration of saline groundwater toward freshwater zones;
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 225

• Acquisition of three dimensional data concerning aquifer structure by well testing


with traces in order to accurately define hydraulic conductivity distribution, integrity
of confining layer at the bottom of the phreatic aquifer or the existence of leaky
confined aquifers in the subsurface.

Water management for this extended coastal area needs to overcome the current
fragmentation of the water authorities and the lack of communication that has affected the
political decisions until the present. This fragmentation has not permitted the understanding
of the most important priorities for a proper integrated management. Choices will have to be
made about how to use the land: do we want to create a wetlands buffer zone between the sea
and land or continue draining to preserve farmland? Do we want to preserve local pine
forests at any cost thereby reducing the quality of the aquifer or is it possible to have more
sustainable vegetation? Do we allow the construction of further tourist establishments that
feed the local economy or is it more important to protect the dunes and beaches as a recharge
area[0]?
We hope that the scientific and integrated overview of the Emilia-Romagna coast and its
issues may contribute to improving the water management in this area and to preserving the
remaining freshwater supplies.

Acknowledgement

We thank the University of Bologna, Autorità dei Bacini Romagnoli, the Emilia-Romagna
Region, the Regional Park of the Po Delta and the Municipality and Province of Ravenna for
funding the several projects that have carried out this chapter.

References
Amorosi, A.; Colangelo, M.L.; Pasini, G.; Preti, D. Sedimentary response to Late Quaternary
sea-level changes in the Romagna coastal plain (Northern Italy). Sedimentology 1999, 46,
99-121.
Amorosi, A; Asioli, A.; Bondesan, M.; Cibin, U; Colangelo, M.L.; Correggiari, A.; Pasini,
G.; Preti, D.; Roveri, M.; Sarti, G.; Severi, P.; Stefani, M.; Trincardi, F.; Vaiani, S.C.;
Vincenzi, S. (2002). Dalle conoidi pedemontani al mare aperto: architettura stratigrafica
tardo-quaternaria dell’Adriatico settentrionale e della pianura costiera emiliano-
romagnola. Le Pianure: conoscenza e salvaguardia. Il contributo delle Scienze della
Terra, Ferrara.
<http://www.regione.emilia-romagna.it/wcm/geologia/canali/geologia/relazioni_scientifiche/
conoidi_pedemontane.htm#2194560>
Antonellini, M.; Minchio, A.; Gabbianelli, G. L’intrusione salina negli Aquiferi Costieri
Emiliano-Romagnoli. In: Pianificazione e tutela del territorio costiero. Questioni,
metodi, esperience a confronto. Ferraci, E. (Ed.); Published by Maggioli Editore, 2007;
pp 105-117.
226 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Antonellini, M.; Mollema, P.; Giambastiani, B.; Bishop, K.; Caruso, L.; Minchio, A.;
Pellegrini, L.; Sabia, M.; Ulazzi, E., Gabbianelli, G. *Saltwater intrusion in the coastal
aquifer of the Southern Po Plain, Italy. Hydrog. J.; published online, June 2008.
Armi, L. The hydraulic of two flowing layers with different densities. J. Fluid Mech. 1986,
163, 27-58.
ARPA (2002). Attivitá della Regione Emilia-Romagna nel campo della difesa costiera.
<http://www.arpa.emr.it/ingamb/difesa_costa.htm>
ARPA (2003). Studio della subsidenza antropica generata dall'estrazione di acqua di falda
lungo la costiera emiliano romagnola. Rapporto tecnico, Servizio Geologico e dei Suoli,
Regione Emilia-Romagna, 2003.
<http://www.regione.emilia-romagna.it/wcm/geologia/canali/subsidenza/progetti_specifici/
02_subsidenza_antropica/Report_ravenna.pdf>
<http://www.regione.emilia-romagna.it/wcm/geologia/canali/subsidenza/progetti_specifici/
02_subsidenza_antropica.htm>
Baines, P.G. Topography effect in stratified flows. Cambridge University Press; 1995; 498 pp.
Bear, J.; Cheng, A.H.-D.; Sorek, S.; Ouazar, D; Herrera, I. Seawater intrusion in coastal
aquifers - Concepts, methods and practices (Theory and application of transport in
porous media); Springer, 1999; 640 pp.
Bondesan, M.; Calderoni, G.; Dal Cin, R. Il litorale delle province di Ferrara e Ravenna (Alto
Adriatico): evoluzione morfologica e distribuzione dei sedimenti. Bol. Soc. Geol. It.
1978, 97, 247-287.
Bondesan, M. Le zone umide salmastre dell’Emilia-Romagna: aspetti geografici e
morfologici. In: Aspetti naturalistici delle zone umide salmastre dell’ Emilia-Romagna;
Corbetta F. Ed; Bologna, IT, 1990; pp 21-69.
Bondesan, M.; Favero, V.; Viñals, M. New evidence on the evolution of the Po delta coastal
plain during the Holocene. Quat. Intern. 1995(a), 29/30, 105-110.
Bondesan, M.; Castiglioni, G.B.; Elmi, C.; Gabbianelli, G.; Marocco, R.; Pirazzoli, P.A.;
Tomasin, A. Coastal areas at risk from surges and sealevel rise in Northeastern Italy. J.
Coastal Res. 1995(b), 11(4), 1355-1379.
Bouwer, H.; Rice, R.C. A slug test for determining hydraulic conductivity of unconfined
aquifer with completely or partially penetrating wells. Water Resour. Res. 1976, 12, 423-
428.
Bouwer, H. The Bouwer and Rice slug test – an update. Ground Water 1989, 27 (3), 304-309.
Brambati, A.; Carbognin, L.; Quaia, T.; Teatini, P.; Tosi, L. The Lagoon of Venice:
geological setting, evolution and land subsidence. Episodes, 2003, 26 (3), 264-268.
Burger, H.R. Exploration geophysics of shallow subsurface. Prentice-Hall, Englewood Cliffs,
NJ, 1992; 489 pp.
Carbognin, L.; Tosi, L.; Teatini, P. Analysis of actual land subsidence in Venice and its
hinterland (Italy). In: Land subsidence – Proc. 5th Int. Symp. On Land Subsidence;
Barends, F.B.J.; Frits, F.J.J. and Schrader, F.H. (Eds), Rotterdam, The Netherland, 1995;
pp 129-137.
Carminati, E.; Doglioni, C.; Scrocca, D. Magnitude and causes of long-term subsidence of the
Po Plain and Venetian Region. In: Flooding and Environmental Challenges for Venice
and its Lagoon: State of Knowledge. Fletcher, C.A.; Spencer, T. Ed., Cambridge
University Press, 2005, pp 21-28.
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 227

Di Dio, G.; Molinari, F.C.; Duroni, G.; Sangallo, A.M.; Zecchi, A. A study of the
groundwater resources in the Ferrara province (Po river basin, Northern Italy). Emilia-
Romagna Cartography Office, S.EL.CA, Firenze, IT, 1999; 120 pp.
<http://www.regione.emilia-romagna.it/wcm/geologia_en/Sections/Water_resources/
rel_scentifiche/ferrara_study.htm#2572892>
Dupuit, J. Estudes théoriques et pratiques sur le mouvement des eaux dans les canaux
découverts et à travers les terrains perméables, 2nd ed.; Dunod,Paris, France, 1863; 304
pp.
EC (European Commission). The EU Water Framework Directive – integrated river basin
management for Europe. Directive 2000/60/EC, Official Journal of the European
Communities (OJ L 327), 2000; 72 pp.
< http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2000:327:0001:0072:EN:
PDF>
Fetter, C.W. Applied Hydrogeology; Fourth Edition; Prentice Hall Ed., Upper Saddle River,
N.J, 2001; 691 pp.
Gambolati, G., Teatini, P. Numerical analysis of land subsidence due to natural compaction of
the Upper Adriatic Sea basin. In: CENAS: Coastal evolution of the Upper Adriatic sea
due to sea level rise and natural and anthropogenic land subsidence; Gambolati, G.
(Ed.), Kluwer Academic Publishing, Series 28, Chapter 5, 1998; pp 103-131.
Giambastiani, B.; Antonellini, M.; Gabbianelli, G. Evaluation of a well-injection recharge
system to contrast saltwater intrusion in a coastal forest: the San Vitale Pinewood,
Ravenna (Italy). In: Proc. 5th European Congress on Regional Geosciences Cartography
and Information System – Earth and Water, Vol. 1; Instituit Cartogràfic de Catalunya and
Instituit Geològic de Catalunya (Eds), Barcelona, Spain, 2006; pp 29-31.
Giambastiani, B.M.S., Antonellini, M.; Oude Essink, G.H.P.; Stuurman R.J. Saltwater
instrusion in the unconfined coastal aquifer of Ravenna (Italy): A numerical model. J.
Hydrol. 2007, 340, 91-104.
Guo, W.; Langevin, C.D. User’s Guide to SEAWAT: A computer program for simulation of
three dimensional variable-density ground water flow: Techniques of Water-Resources
Investigations, Book 6, Chapter A7; U.S.G.S; 2002; 79 pp.
< http://fl.water.usgs.gov/publications/Abstracts/twri_6_A7_guo_langevin.html>
Hubbert, M.K. The theory of ground-water motion. J. Geol. 1940, 48, 785-944.
IAH. Guiding Principles for Sustainable Groundwater Management. IAH (International
Association of Hydrogeologists, 2004.
< http://www.iah.org.au/pdfs/SY%20Posn%20Paper.pdf>
Idroser. Progetto di piano per la difesa dal mare e la riqualificazione ambientale del litorale
della Regione Emilia-Romagna – Relazione Generale. Bologna, IT, 1996; 365 pp.
IPCC. Fourth Assessment Report: Climate Change 2007. Intergovernmental Panel on Climate
Change, UNEP, 2007.
< http://www.ipcc.ch/ipccreports/ar4-wg1.htm>
< http://www.ipcc.ch/ipccreports/ar4-wg2.htm>
< http://www.ipcc.ch/ipccreports/ar4-wg3.htm>
Jacob, C.E. Flow in groundwater. In: Engineering Hydraulics; Rouse, H. Ed.; John Wiley,
NY, 1950; 321-386.
228 Beatrice M.S. Giambastiani, Pauline N. Mollema and Marco Antonellini

Loke, M.H. Electrical imaging surveys for environmental and engineering studies - A
practical guide to 2-D and 3-D surveys. RES2DINV Manual, IRIS Instruments, 2001; 67
pp.
< http://www.terrajp.co.jp/lokenote.pdf>
Marchesini, L.; Amorosi, A.; Cibin, U.; Zuffa, A.; Spadafora, E.; Preti, D. Sand composition
and sedimentary evolution of a Late Quaternary depositional sequence, Northwestern
Adriatic Coast, Italy. J. Sedim. Research 2000, 70 (4), 829-838.
Minchio, A.; Antonellini, M.; Caruso, L.; Gabbianelli, G.; Giambastiani, B.; Laghi, M.;
Stecchi, F. The role of dunes in coastal hydrology and integrated coastal zone
management. In: Epitome, Vol. 2, Proc. GEO ITALIA conference, Publisher: FIST, 2007.
Mollema, P.; Giambastiani, B.; Ulazzi, E.; Gabbianelli, G. The influence of climate change
on the water budget in the Regional Park of the Po River Delta. In: Special Issue
MEDCOAST 03: Climatic change and foreseeable impacts on deltas: a new challenge for
the conservation of protected areas, Proc. 6th Int. Conf. on the Mediterranean Coastal
Environmental, Vol. 4; Gabbianelli, G. & Sangiorgi F. (Ed.s), 2003; pp 69-79.
Mollema, P.; Antonellini, M.; Minchio, A.; Gabbianelli, G. The Influence of Three-
dimensional Dune Topography on Salt Water Intrusion in Marina Romea, Italy: A
Numerical Modeling Study Using LIDAR Data. In: Proc. SWIM20 – 20th Salt Water
Intrusion Meeting; June 23-27, 2008; Naples, Florida; pp 151-154.
Oude Essink, G.H.P. MOC3D adapted to simulate 3D density-dependent groundwater flow.
In: Proc. MODFLOW’98 Conf.; Golden, Colorado, USA, 1998; pp 291-303.
Oude Essink, G.H.P. Simulating density dependent groundwater flow: the adapted MOC3D.
In: Proc. 15th Salt Water Intrusion Meeting; Ghent, Belgium, 1999; pp 69-79.
Oude Essink, G.H.P. Salt Water Intrusion in a three-dimensional groundwater system in The
Netherland: a numerical study. Transport Porous Med. 2001, 43 (1), 137-158.
Oude Essink, G.H.P. Modeling Three-Dimensional Density Dependent Groundwater flow at
the Island of Texel, The Netherland. In: Coastal Aquifer Management: monitoring
modeling and case studies, Cheng, A.H-D. and Ouazar, D. (Eds); Lewis Publishers, 2004;
pp 77-94.
Pieri, M.; Groppi, G. Subsurface geological structure of the Po Plain, Italy. In: Progetto
finalizzato Geodinamica, Quad. CNR, 414, Roma, IT, 1981; pp 1-23.
Preti, M. Eustatismo, subsidenza e linee di intervento per la difesa del territorio costiero in
Emilia-Romagna. In: Mare e cambiamenti globali, ICRAM, ARPA – Ingegneria
Ambientale, Bologna, IT, 2000, pp 167-179.
Ranjan, S. P.; Kazama, S.; Sawamoto, M. Effects of climate and land use changes on
groundwater resources in coastal aquifers. J. Environ. Manage. 2006, 80, 25-35.
RER Regione Emilia-Romagna. Opere di difesa della costa – Sistemazione della foce del
fiume Lamone mediante dragaggio, sistemazione del molo sinistro e ripascimento verso
Marina Romea in Comune di Ravenna - Progetto esecutivo, Studio di incidenza
ambientale; Servizio Tecnico Bacino Fiumi Romagnoli, Ravenna, Italy, 2006.
Regione Emilia-Romagna & ENI-AGIP S.P.A. Riserve idriche sotterranee nella Regione
Emilia-Romagna. Di Dio, G. Ed. Florence, IT, 1998; pp 1-120.
<http://www.regione.emilia-
romagna.it/wcm/geologia/canali/acque/rel_scientif/riserve_idriche_sotterr_01.htm>
Rizzini, A. Holocene sedimentary cycle and heavy mineral distribution, Romagna-Marche
coastal plain, Italy. Sedim. Geol 1974, 11, 17-37.
Groundwater Management in the Northern Adriatic Coast (Ravenna, Italy)… 229

Sabia, M.; Antonellini, M.; Gabbianelli, G.; Giambastiani, B.; Lapenna, V.; Perrone, A.;
Rizzo, E. Indagini geoelettriche per lo studio di strutture idrogeologiche nella pineta di
San Vitale (RA). In: Atti del Convegno Nazionale di Geofisica della Terra Solida; Roma,
IT, 2005; pp 357-360.
Schijf, J.B.; Schonfeld, J.C. Theoretical consideration on the motion of salt and fresh water.
In: Proc.. Minnesota. Intern. Hydraulics Convention; Minneapolis, Minnesota, 1953; pp
321-333.
Selli, R.; Ciabatti, M.. L’abbassamento del suolo della zona litoranea ravennate. G. Geol.
1977, 42 (1), 1–47.
Simeoni, U., Atzeni, P.; Bonora, N.; Borasio, E.; Del Grande, C.; Gabbianelli, G., Gonella,
M.; Tessari, U.; Valpreda, E.; Zamariolo, A. Integrated management study of Comacchio
Coast (Italy). J. Coastal Res. 2002, Special Issue 36, 686-693.
Starr, M. Watbal: A model for estimating monthly water balance components, including soil
water fluxes. In: 8th Annual Report 1999, UN ECE ICP Integrated Monitoring, The
Finnish Environment 325; Kleemola, S. & Forsius, M. (Eds.), Helsinki, Finland, 1999; pp
31-35.
<http://www.metla.fi/hanke/3098/ewat_bal.htm>
Teatini, P.; Ferronato, M.; Gambolati, G.; Gonella, M. Groundwater pumping and land
subsidence in the Emilia-Romagna coastland, Italy: Modeling the past occurrence and the
future trend. Water Resour. Res. 2006, 42, 1-19.
Tsoar, H. Sand dunes mobility and stability in relation to climate. Physica A 2005, 357 (1),
50-56.
Ulazzi, E.; Antonellini, A.; Gabbianelli, G. Saltwater intrusion in a unconfined coastal
aquifer: the case study of Cervia (North Adriatic Sea, Italy). In: Integrated Water
Management: Practical Experiences and Case Studies; P. Meire et al. (Eds), Volume 80;
2008; pp 295-308.
In: Groundwater: Modelling, Management… ISBN: 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 231-257 © 2008 Nova Science Publishers, Inc.

Chapter 9

HYDROGEOLOGY AND GEOCHEMISTRY


OF THE FRIULI VENEZIA GIULIA PLAIN ALLUVIAL
AQUIFERS, NORTHEASTERN ITALY

Franco Cucchi, Giuliana Franceschini* and Luca Zini


Department of Geological, Environmental and Marine Sciences,
University of Trieste, Italy

Abstract
The FVG Plain has been formed by river systems (mainly Tagliamento and Isonzo) as
well as deposition and reworking of marine and terrestrial deposits. From a geomorphological
and hydrogeological perspective, the FVG Plain consists of two provinces separated by a
resurgence belt that covers a strip 2 to 8 km wide and 80 km long. The Upper Friuli Plain,
composed mostly of calcareous and dolomitic gravels and characterised by the lack of surface
drainage waters hosts an unconfined aquifer (up to 150 m thick). The Lower Friulian Plain is
characterised by multi-layer confined aquifers. A detailed reconstruction of the
hydrogeological setting has indicated that there are a maximum of ten confined aquifers in the
southwestern part of the FVG Plain, whereas only 6 confined aquifers are recognised in the
southeastern part due to tectonic thrusts. However, considering their chemical and physical
characteristics these groundwater layers are subdivided in shallow and deep confined aquifers.
The shallow and deep confined aquifers have groundwater in the temperature range of 10 to
20°C. The shallow and deep confined aquifers are separated by a 10 to 15 m thick
impermeable layer of silty material at approximately 100 to 110 m depth. The thickness of the
multi-layered confined aquifer increases towards the Adriatic Sea where it reaches a thickness
of up to 500 m. Thermal groundwaters (up to 60°C) have been found in the southern part of
the FVG Plain at depths of 350 m.
The pattern of geochemical and stable isotope variations suggests that the unconfined and
shallow confined groundwaters are recharged mostly by rainfall and local river (mainly
Tagliamento) infiltrations. This fast recharge process makes these groundwaters susceptible to
contamination to contamination. Four hydrogeological provinces have been recognised for
these subsurface groundwaters. Radiocarbon dating indicates that there is very little continuity
between the subsurface groundwaters and the deeper aquifers that have complex groundwater
circulations and have substantially changed during the varying temperature regimes of the

*
E-mail address: giuliana.franceschini@units.it
232 Franco Cucchi, Giuliana Franceschini and Luca Zini

Holocene-Pleistocene. Significant late Quaternary sea-level fluctuations, associated with


alternating cooler and wetter periods, would have changed the hydraulic gradients and
partially or completely disconnected the deeper parts of the aquifer systems from the more
active surface circulations. Comparison with deep confined aquifers in other regions of the
Padain Plain indicates that the recharge rates of these deep confined aquifers are low and that,
consequently, the deep groundwaters are very sensible to overexploitation.

Introduction
Although the Friuli Venezia Giulia Plain (also known as the Friuli Plain) geographically is
the eastern sector of the Padain Plain it is characterised by different geohydrogeological
developments (Figure 1). In fact, the Padain Plain has been formed by the Po River, the
largest Italian river and one of the tenth largest European rivers, whereas the Friuli Plain has
been shaped by the Tagliamento River and other minor regional rivers which have never been
tributaries of the Po River. Therefore, even if the Friuli Plain is considered part of the Padain
Plain, geologically and geomorphologically the two plains have distinct characteristics. In
addition to this, the Friuli Plain has the peculiarity of showing a very short distance between
the Alps and the coastline averaging 40 km, and getting even narrower up to disappear at the
contact with the karst area.

Figure 1. Map of the northern Italy.


Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 233

Figure 2. Map of the Friuli Venezia Giulia with average rainfall values in mm/yr.

The Friuli Plain is bordered to the north by the Carnic Alps, to the east by the Julian Alps
and the Classical Karst and the south border is formed by the Adriatic Sea (Figure 1).
Towards the west the Livenza River marks the border between the Friuli Plain and the
Venetian Plain. The Friuli Plain has several climatic cells due to it extend from the Alps to the
Adriatic Sea and one of the highest rainfalls in Europe. In fact, in the northeaster sector the
average rainfall is 3200 mm/yr (Figure 2). In the plain the average rainfall is 1200 mm/yr
lowering down to 1000 mm/yr in the coastal areas. Most of the rainfall is concentrated in the
autumn and in the spring seasons (average monthly rainfall in November is 120 mm and 105
mm in June). The average temperature of the plain is of 13.5 °C with a maximum of 23°C in
July and a minimum of 2.6°C in February. Approximately 1 million people live in the area,
with 1/5 of the population located in the territorial capital city of Udine.

Geological and Geomorphological Framework


The Friuli area (Figure 3) is characterized by sedimentary rocks ranging from Paleozoic to
Quaternary (Slejko et al., 1989; Bosellini, 2004; Carulli, 2006; Fontana, 2006). The Paleozoic
rocks that outcrop in the northern part of the area are mainly composed of terrigenous rocks,
limestones, frequent evaporites and minor volcanic deposits. The geologic units of the central
part of the area are mainly made up of limestones and carbonatic rocks (Triassic–Cretaceous),
234 Franco Cucchi, Giuliana Franceschini and Luca Zini

belonging to the thick shelf complex of the Friuli Platform. This Mesozoic platform is in
average 750 to 1000 m deep and it is overlain by flysch and molasses that represent the
Cenozoic deposits. Pliocene-Quaternary alluvial deposits are found above the flysch and
molasses (Cati et al., 1987; Carulli, 2006).

Figure 3. Location of the study area (dotted line= Resurgence belt, P = Pluviometer,
crosses = phreatic well samples, triangles = shallow confined well samples and circles =
deep confined well samples; 1 = Cellina River, 2 = Livenza River, 3 = Tagliamento
River, 4 = Torre River, 5 = Natisone River and 6 = Isonzo River).

The main surface drainage of the Friuli Plain is the Tagliamento River that is 180 km
long and is the 6th longer Italian river. Its headquarters are located in the Carnic Alps and it
has a catchment of 2871 km2 (Basin Authority for the North Adriatic Rivers, 1997). The
average discharge of the Tagliamento River, 30 km inland from the Adriatic Sea, is 90-100
m3/sec (Fontana and Bondesan, 2006), with a maximum discharge of 3,000 to 4,500 m3/sec
and over 6,000 m3/sec over a return period of 500 years (Maione and Machne, 1982;
Foramitti, 1990; Spalviero, 2003). The ratio between average and maximum values ranges
from 1/40 to 1/60, much higher than the other major European rivers (such as Po, Danube
with values of 1/20). The great difference between average and maximum discharge of the
Tagliamento River makes it having a torrential behaviour which creates frequent natural
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 235

changes to the river banks. For this reason, the Tagliamento River is considered the only river
that still retains its natural or semi-natural river characteristics in the entire Alps (Spalviero,
2003). In the Upper Fiulian Plain the Tagliamento River flows in the gravel deposits in a 1.5
km wide bed. The average discharge of the Tagliamento River in the Upper Friulian Plain is
10-15 m3/sec and null during summer months. After the resurgence belt, the river bed is only
50 m wide and fully braided with high bed load and associated relatively steep gradient
(average depth of 10-15 m). The lower part of the river is a single channel showing semi-
meandering pattern and from Latisana (25 km from the coast) the river flow is under tidal
breakwater effect. The lower part of the river has been heavily canalised to provide water to
local agricultural fields and to protect local villages from flooding.
The second largest river in Friuli Venezia Giulia is the Isonzo River (known as Soča in
Slovenia). The Isonzo River has a pronounced mountainous character with an average
elevation of about 600 m above sea level. Its average discharge is 172 m3/sec but maximum
discharge of 4,400 m3/sec over the period of 1925-1953. Other perennial rivers in the Friuli
Venezia Giulia Region (Basin Authority for the North Adriatic Rivers, 1997) are the Torre
River (average discharge of 17 m3/sec), the Natisone River (7 m3/sec), the Cellina River
(average discharge of 16 m3/sec) and the Meduna River (average discharge of 11 m3/sec). The
Livenza River (average discharge of 85 m3/sec) is the largest of the resurgence rivers in the
Friulian Plain. The Livenza River has three perennial karst springs called Molinetto,
Santissima and Gorgazzo characterised by a total discharge of 15 m3/sec.

Aquifer Descriptions
The shape, dimensions and groundwater capacities of the aquifers, aquitards and aquicludes
of the Friuli Plain have been mostly described in the 70’s. The knowledge has been
continuously updated using geological information, stratigraphical correlations from water
exploration wells, seismic profiles and deep wells created for water, oil and gas explorations
(Stefanini and Cucchi 1976, 1977; Mosetti, 1983; Cucchi et al., 1998, 1999, 2000; Carulli,
2006). The quality of the data used to build the geohydrological model and therefore the
understanding of the groundwater characteristics of the hosting layers has not always been of
a great quality mostly because the data used to create the Friuli hydrological model have been
acquired for different purposes. For instance, the seismic data has been taken for oil and gas
exploration and therefore to delineate the thickness and composition of the lower consolidated
Mesozoic carbonate rocks. In addition, filter positions of water wells have not always been
recorded by local drillers or the wells have not been well cemented and groundwaters from
different aquifers have been mixed. However, the great quantity (the stratigraphical database
used for this work has 2000 records and the chemical database approximately 4000 analyses)
of data available allows to describe the geohydrological situation of the plain fairly well.
From a morphological point of view, the Friuli Plain is divided into two units, the Upper
Friulian and the Lower Friulian, which are separated by the resurgence belt (Figure 4). The
Upper Friulian Plain is formed by thick fluvial and glacial deposits, coarse to very coarse in
size (gravels with sandy layers and locally with lenses of siltly impermeable materials). The
coarse materials are loose and become cemented (conglomerates and breccias) with depth.
236 Franco Cucchi, Giuliana Franceschini and Luca Zini

Figure 4. Idealized north-south cross section of the entire length of the Friuli Venezia Giulia that
highlights the High and Low Plains.

Figure 5. West-East- cross section of the Lower Friulian Plain showing the ten confined aquifers
(labelled as A, B, C, D, E, F, G, H, I, L).

The permeability of the loose gravely deposits is extremely high. The gravel clasts are made
of carbonatic material and the silica fraction increases eastwards due to the thick flysch
deposits of the eastern portion of the Friuli Venezia Giulia Region. The Mesozoic bedrock
underlay the unconfined (U) aquifer hosted by the Upper Friulian Plain. This carbonatic shelf
thickens up southward. Westward, thrusts duplicate the sequence. In the eastern part of the
region, the thrust system is shallower and therefore the alluvial deposits are thinner (Figure
4). The unconfined aquifer may be considered a single aquifer with the water table surface
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 237

orientated approximately in WNW to ESE direction and hydraulic gradients of 8*10-3 in the
northern part lowering down to 1-2*10-3 in proximity of the resurgence belt (Martelli et al.,
2007). A series of flow direction measures collected using thermo-resistivity data have
demonstrated that the unconfined aquifer is continuous but marked by hydraulic
transmissivity changes. The difference between the maximum and minimum groundwater
levels can reach up to 40 m depending on the seasons (Cucchi et al., 1999; Martelli et al.,
2007). The unconfined groundwater is fed mainly by infiltrations of river waters, by rainfall
from the pre-Alps areas and also by local rainfall (Cucchi et al., 2007; in press). In addition,
an important groundwater source is provided by the mountainous deep aquifers although it is
presently impossible to quantify the water contained in these sedimentary aquifers, with
karstic or fractured fissures. The Upper Friulian Plain is rich either in superficial and
groundwaters.

Figure 6. South-North cross section with the ten confined aquifers in the southern part and five in the
northern part close to the resurgence belt.

The Lower Friulian Plain is characterized by thicker impermeable silty deposits. The
plain slope is close to being sub horizontal and as a consequence large coastal areas are below
sea level. From the resurgence belt southwards, the silt and clay layers are more and more
frequent and their thickness increases from few centimeters (in the Upper Friulian Plain) to
several meters (in the western Lower Plain up to 35 m of continuous clay layers have been
mapped). Close to the resurgence belt, the silty-clay layers become better developed, thicker
and they are laterally continuous creating a thick impermeable layer found in the entire length
of the plain. The groundwater in this area is hosted first by aquicludes and then by true
aquifers. A series of data such as private and municipal well drilling, seismic data,
stratigraphic correlations have demonstrated that from the resurgence belt to the Adriatic Sea
238 Franco Cucchi, Giuliana Franceschini and Luca Zini

there are 10 confined aquifers that represent very important and large groundwater sources for
the entire Friuli Venezia Giulia region. These ten confined aquifers are found in the Friuli
Plain located between the Tagliamento and the Livenza rivers (Figures 5, 6). Due to Alpine
thrusting, between the Tagliamento and the Isonzo rivers only 7 aquifers have been
recognized. In the area of these confined aquifers, the superficial deposits are mostly
composed of silty sands and peat deposits. The gravelly covering material is not well
developed and hosts a discontinuous perched unconfined aquifer.
A reconstruction of the aquifer layers in the Lower Friuli Plain (Stefanini and Cucchi,
1976, 1977, Cucchi et al., 1999; Granati et al. 2000, 2001) revealed the presence of ten
confined layers (labelled as A, B, C, D, E, F, G, H, I, L in figures 5, 6). The calculated
thickness and depths of layer H, I, L are indicative and mostly taken from seismic data and
few water exploration drilling data. The average depth of layer A increases in a southward
direction to 80 m bsl and the thickness reaches a minumum of 30 m near the resurgence belt.
The greatest average depths of layer B (120 m bsl) is found in the central part of the Lower
Friulian Plain in the vicinity of the resurgence belt. The average thickness of this layer is
about 8 m. The average depth of layer C close to the resurgence belt is 130 m bsl, with an
average thickness of 7 m. Layer D is found at depths of 160 m bsl and has an average
thickness of 11 m, whereas layer E is located at the average depth of 200 m bsl and has an
average thickness of 18 m. Layer F is encountered at an average depth of 250 m bsl and its
thickness reaches a value of 40 m in the vicinity of the Tagliamento River. The remaining
aquifer layers have been found only on the part of the plain between the Tagliamento and
Isonzo rivers. Layer G is found at depths of 270 m and this layer is in average 11 m thick.

Figure 7. Reconstruction of the chemical characteristics of the confined aquifers.

This layer is characterised by low thermal waters with temperature reaching values of
34°C. Layer H is the first aquifer that contains thermal waters and has been found at and
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 239

average depth of around 400 m and is 33 m thick. Layer H has an average water temperature
of 45°C. Very little information is available for the thermal aquifers found at 460 m (layer I)
and 550 m (layer L) that are contained in sandy layers and have thermal water up to 60°C. A
reconstruction of the chemical characteristics of the confined aquifers is presented in Figure 7
(after Cucchi et al., 1999).
Accordingly to their chemical and physical characteristics, the multilayer aquifers may be
subdivided into three different groundwater types: shallow confined (layers A and B) and
deep confined (layers C, D, E, F) groundwaters with water temperature ranging from 10 to
20°C and thermal deep confined groundwater with temperatures up to 60°C (layers G, H, I,
L). The shallow and deep confined groundwaters are separated by a 10 to 35 m continuous
thick impermeable layer found at depths of 120 to 130 m, which separates the B and C layers.
This low permeability layer (hydraulic conductivity of 10-7m s-1, Mosetti, 1983) has been also
reported in several water wells of the Po Plain area (Pilla, 1998; Pilli, 2005). The deep
confined multilayer aquifer has water temperature ranging from 15° C to approximately 25°C
up to 450-500 m depending on geological conditions of the area. Thermal aquifers have been
encountered at depths greater then 500 m and are used by local industries as well as for
domestic usages. The presence of geothermal groundwater in the southern part of the Lower
Friulian Plain is fed by saline water (at an estimated temperature of up to 60 °C from the
Mezosoic carbonate reservoir which is interested by a positive geothermical gradient
(Regione Autonoma Friuli-Venezia Giulia, Direzione Centrale Ambiente e Lavori Pubblici,
2007).

Groundwater Depletion
Groundwater depletion is the inevitable and natural consequence of withdrawing water from
an aquifer. In the past half-century, ready access to the unconfined and confined groundwater
in the Friuli Plain for industrial, municipal and agricultural supplies has resulted in water
quality deterioration, rising of pumping costs, to the extent that well yields have decrease. As
depletion continues, its impacts worsen, also because of the absence of a specific water
supply legislation. In addition, the abundance of superficial waters in Friuli Venezia Giulia
has led local population to not consider water as a resource to be managed and preserved.
Decentralized management by local governmental agencies has resulted in a lack of
coordination between surface and groundwater use, despite their vital physical connections
and this will likely increase in groundwater depletion in the next decades. In the southern
sector of the Friulian Plain, Martelli and Granati (2007) have estimated that the active
recharge is 820 millions m3, whereas the total water withdrawal is 701 millions m3.
Therefore, the active recharge may be compared almost to the estimated groundwater
extraction from the Lower Plain’s aquifer systems. This unnatural and dangerous situation is
also reflected in the continuous losing in pressure of the shallow confined aquifers that have
lost their artesian behaviour.
Table 1. Chemical and isotopical composition of groundwater samples from the Friuli Venezia Giulia Plain.
δ18O Std.
Borehole ID Type Elevation Depth nch EC pH TDS Alk TH SAR Ca2+ Mg2+ Na+ K+ Cl- HCO3 - SO42- NO3- Water Type SIdol SIcal SIgyp SIani SIhal nis δ18O
dev.
(‰VSMO
(m a.s.l.) (m) (µS/cm) (mg/L) (mg/L) (mg/L) (meq/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)
W)
637 U 3 2 6 606 7.2 590 328 371 0.12 89.4 35.9 5.2 0.9 10.3 399 48.3 0.2 Ca-Mg-HCO3 0.08 0.14 -1.88 -2.13 -8.84 2 -7.30 0.03
255 U 19 12 12 597 7.4 569 316 348 0.09 105.2 20.7 4.0 1.4 8.2 386 17.3 23.3 Ca-Mg-HCO3 0.23 0.36 -2.23 -2.48 -9.06 2 -6.81 0.06
204 U 40 15 11 563 7.4 506 211 329 0.11 89.3 25.7 4.4 0.9 7.0 257 106.0 14.1 Ca-Mg-HCO3-SO4 -0.16 0.10 -1.51 -1.76 -9.07 1 -8.44 -
236 U 82 15 15 489 7.6 441 163 286 0.08 82.4 19.4 3.2 0.9 3.8 199 124.6 2.4 Ca-Mg-HCO3-SO4 -0.16 0.14 -1.45 -1.70 -9.48 1 -8.63 -
449 U 61 15 7 936 6.9 928 535 562 0.20 149.6 45.8 10.7 0.7 17.1 652 17.6 33.4 Ca-Mg-HCO3 0.12 0.21 -2.18 -2.43 -8.33 - - -
610 U 20 15 8 605 7.4 556 265 343 0.14 87.2 30.4 6.1 1.8 11.2 324 53.9 42.3 Ca-Mg-HCO3 0.14 0.19 -1.83 -2.07 -8.73 2 -6.80 0.13
613 U 30 17 8 626 7.3 578 268 366 0.09 94.9 31.2 4.0 1.1 8.7 327 72.7 38.0 Ca-Mg-HCO3-SO4 0.04 0.16 -1.67 -1.92 -9.03 2 -7.61 0.05
609 U 24 18 8 631 7.4 584 294 361 0.12 94.5 30.3 5.4 1.7 10.4 359 38.6 44.4 Ca-Mg-HCO3 0.21 0.25 -1.94 -2.19 -8.85 2 -7.33 0.03
612 U 21 19 7 607 7.4 556 270 346 0.15 88.2 30.6 6.3 1.0 12.3 329 39.4 48.7 Ca-Mg-HCO3 0.04 0.15 -1.95 -2.20 -8.68 2 -6.82 0.04
611 U 20 20 8 651 7.3 607 299 375 0.14 99.2 31.0 6.3 2.0 12.4 365 48.4 43.0 Ca-Mg-HCO3 0.16 0.23 -1.83 -2.08 -8.67 2 -7.01 0.02
3776 U 51 20 9 588 7.4 568 303 352 0.07 97.4 26.5 3.2 1.1 8.0 370 13.6 47.8 Ca-Mg-HCO3 0.16 0.25 -2.38 -2.63 -9.18 - - -
31126 U 28 20 3 260 - 212 115 148 0.07 44.5 9.0 2.0 0.6 2.8 140 7.4 5.2 Ca-Mg-HCO3 -1.78 -0.68 -2.81 -3.05 -9.81 1 -8.55 -
1524 U 37 22 9 505 7.6 480 261 303 0.06 81.2 24.4 2.5 0.7 5.8 319 14.5 32.2 Ca-Mg-HCO3 0.20 0.26 -2.39 -2.64 -9.40 2 -7.66 0.06
1357 U 34 23 12 348 7.9 325 186 200 0.05 49.5 18.5 1.6 0.3 3.6 227 7.1 15.2 Ca-Mg-HCO3 0.17 0.21 -2.83 -3.08 -9.94 2 -8.09 0.01
624 U 60 25 5 506 7.5 447 178 289 0.09 77.5 23.3 3.5 0.9 5.3 218 107.0 11.5 Ca-Mg-HCO3-SO4 -0.13 0.10 -1.54 -1.79 -9.30 1 -8.64 -
639 U 25 25 8 639 7.3 592 310 373 0.09 101.6 28.9 4.1 1.7 12.2 378 24.5 40.7 Ca-Mg-HCO3 -0.02 0.17 -2.11 -2.36 -8.88 1 -7.02 -
1298 U 17 25 14 589 7.4 567 302 353 0.07 95.7 27.6 3.3 1.0 9.3 369 24.7 34.7 Ca-Mg-HCO3 0.18 0.26 -2.12 -2.37 -9.09 2 -6.84 0.04
1305 U 16 25 14 557 7.4 529 272 326 0.13 83.1 28.9 6.4 1.0 10.6 332 32.9 32.1 Ca-Mg-HCO3 0.15 0.20 -2.04 -2.29 -8.73 2 -7.19 0.01
234 U 141 26 17 548 7.4 518 288 321 0.07 92.4 21.8 2.8 1.1 7.2 351 18.4 19.4 Ca-Mg-HCO3 0.08 0.27 -2.24 -2.49 -9.27 2 -7.03 0.06
211 U 46 27 11 506 7.5 452 200 289 0.11 80.7 21.3 4.3 1.0 5.9 244 86.2 7.5 Ca-Mg-HCO3-SO4 -0.33 0.04 -1.61 -1.86 -9.15 1 -8.64 -
31055 U 32 27 10 298 7.9 265 137 167 0.07 50.4 10.1 2.2 0.9 3.4 167 6.2 6.4 Ca-Mg-HCO3 -1.73 -0.60 -2.86 -3.11 -9.67 2 -8.43 0.11
43 U 41 30 7 495 7.4 470 256 292 0.09 81.9 21.3 3.5 1.2 6.7 312 16.1 27.9 Ca-Mg-HCO3 -0.02 0.19 -2.33 -2.58 -9.22 2 -6.80 0.05
202 U 41 31 12 484 7.6 425 167 278 0.09 76.6 21.1 3.4 0.8 4.6 203 106.6 7.3 Ca-Mg-HCO3-SO4 -0.27 0.06 -1.54 -1.79 -9.37 1 -8.59 -
206 U 33 31 11 589 7.4 537 245 346 0.08 90.2 29.3 3.3 0.8 8.1 299 76.2 29.3 Ca-Mg-HCO3-SO4 -0.08 0.11 -1.65 -1.91 -9.13 1 -7.87 -
31107 U 9 31 12 295 7.7 251 138 166 0.11 47.7 11.3 3.4 0.6 7.2 168 7.0 5.4 Ca-Mg-HCO3 -1.58 -0.57 -2.83 -3.08 -9.27 - - -
219 U 27 32 11 643 7.3 591 296 371 0.14 98.4 30.3 6.3 1.8 11.5 360 48.3 32.4 Ca-Mg-HCO3 0.04 0.18 -1.83 -2.08 -8.71 1 -7.23 -
209 U 50 33 11 604 7.4 546 235 348 0.14 96.0 26.3 5.8 0.9 9.3 286 100.6 19.7 Ca-Mg-HCO3-SO4 -0.10 0.14 -1.51 -1.76 -8.86 1 -8.45 -
212 U 44 33 11 585 7.4 527 237 341 0.08 89.7 28.4 3.3 0.8 8.0 289 79.2 26.5 Ca-Mg-HCO3-SO4 -0.15 0.09 -1.64 -1.89 -9.14 1 -8.04 -
588 U 111 33 5 617 7.2 575 317 357 0.15 104.6 23.1 6.5 0.9 4.8 386 14.5 34.8 Ca-Mg-HCO3 -0.16 0.15 -2.31 -2.56 -9.10 2 -6.25 0.11
216 U 38 35 11 619 7.3 573 288 366 0.08 95.2 31.1 3.5 0.8 9.2 351 44.8 34.8 Ca-Mg-HCO3 -0.12 0.10 -1.87 -2.12 -9.05 1 -7.48 -
638 U 24 35 8 643 7.1 600 330 368 0.10 108.7 23.4 4.4 0.8 15.9 402 13.3 31.6 Ca-Mg-HCO3 -0.35 0.07 -2.34 -2.59 -8.74 1 -6.52 -
205 U 37 36 11 549 7.4 495 213 321 0.08 85.8 25.9 3.4 0.8 6.6 259 93.5 18.4 Ca-Mg-HCO3-SO4 -0.11 0.12 -1.57 -1.83 -9.21 1 -8.33 -
1303 U 21 36 13 663 7.3 621 300 386 0.12 102.8 31.2 6.1 1.5 10.5 366 53.7 47.9 Ca-Mg-HCO3 -0.02 0.15 -1.78 -2.03 -8.73 2 -7.12 0.00
203 U 47 37 11 491 7.6 427 162 280 0.13 77.8 20.8 5.1 0.8 5.6 197 110.8 7.4 Ca-Mg-HCO3-SO4 -0.21 0.09 -1.52 -1.77 -9.11 1 -8.73 -
208 U 24 38 11 682 7.2 636 308 403 0.14 107.0 33.0 6.6 1.8 11.8 376 57.4 41.1 Ca-Mg-HCO3 0.03 0.17 -1.74 -1.99 -8.69 1 -7.34 -
213 U 44 40 11 588 7.3 535 249 345 0.09 91.8 28.1 3.6 1.0 7.4 304 71.0 27.0 Ca-Mg-HCO3-SO4 -0.13 0.11 -1.67 -1.93 -9.13 1 -7.70 -
246 U 47 40 17 477 7.5 455 254 273 0.12 86.4 13.9 4.7 2.1 6.3 310 15.3 11.9 Ca-HCO3 0.06 0.33 -2.31 -2.56 -9.09 2 -6.76 0.06
1322 U 18 40 12 414 7.6 395 223 239 0.07 66.9 17.4 3.2 1.2 5.0 272 11.8 15.7 Ca-Mg-HCO3 0.19 0.28 -2.54 -2.79 -9.45 1 -8.19 -
1369 U 85 40 9 443 7.8 378 138 254 0.07 73.8 16.9 2.5 0.7 2.9 168 108.7 4.6 Ca-Mg-HCO3-SO4 -0.33 0.07 -1.53 -1.79 -9.70 2 -8.90 0.21
31012 U 18 40 4 367 - 306 163 214 0.09 62.9 13.8 2.9 1.4 4.4 199 9.3 12.4 Ca-Mg-HCO3 -1.37 -0.47 -2.61 -2.86 -9.45 - - -
1523 U 42 43 9 408 7.8 381 215 241 0.04 62.5 20.7 1.3 0.4 4.4 262 8.4 21.6 Ca-Mg-HCO3 0.10 0.19 -2.69 -2.94 -9.79 2 -8.03 0.05
Table 1. Continued
δ18O Std.
Borehole ID Type Elevation Depth nch EC pH TDS Alk TH SAR Ca2+ Mg2+ Na+ K+ Cl- HCO3 - SO42- NO3- Water Type SIdol SIcal SIgyp SIani SIhal nis δ18O
dev.
(‰VSMO
(m a.s.l.) (m) (µS/cm) (mg/L) (mg/L) (mg/L) (meq/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)
W)
218 U 50 44 10 566 7.3 509 254 319 0.18 81.8 27.8 7.4 1.1 12.4 309 43.8 24.1 Ca-Mg-HCO3 -0.21 0.04 -1.91 -2.17 -8.60 1 -7.34 -
252 U 35 45 18 274 7.8 253 145 152 0.07 47.6 7.9 2.0 0.8 2.1 177 7.3 4.0 Ca-Mg-HCO3 -0.16 0.20 -2.80 -3.04 -9.92 2 -7.59 0.49
269 U 55 45 30 599 7.3 592 315 357 0.11 91.3 31.2 4.8 1.6 7.4 384 30.2 37.2 Ca-Mg-HCO3 0.21 0.22 -2.07 -2.31 -9.02 2 -6.74 0.08
618 U 20 48 8 581 7.4 538 270 329 0.20 83.4 29.2 8.2 1.9 13.5 329 39.3 33.6 Ca-Mg-HCO3 0.08 0.17 -1.97 -2.22 -8.52 2 -7.19 0.11
1348 U 36 48 8 375 7.9 347 196 219 0.03 55.4 19.7 1.2 0.3 3.8 238 8.9 19.1 Ca-Mg-HCO3 0.09 0.17 -2.70 -2.95 -9.91 2 -7.35 0.28
463 U 25 50 8 510 7.6 474 246 291 0.12 80.8 21.7 4.8 2.0 7.8 300 41.3 16.0 Ca-Mg-HCO3 0.33 0.35 -1.93 -2.18 -8.98 2 -7.24 0.05
633 U 59 50 8 450 7.6 412 221 257 0.13 64.8 23.0 4.7 1.4 7.2 270 20.4 20.9 Ca-Mg-HCO3 0.08 0.17 -2.30 -2.55 -9.04 - - -
634 U 59 50 7 552 7.5 492 243 300 0.27 77.2 26.1 10.8 0.9 15.1 297 43.2 21.4 Ca-Mg-HCO3 0.11 0.19 -1.96 -2.21 -8.35 - - -
1527 U 54 50 9 456 7.7 409 228 260 0.04 66.2 23.1 1.6 0.3 5.1 277 10.1 24.8 Ca-Mg-HCO3 0.07 0.18 -2.59 -2.85 -9.63 2 -7.74 0.09
31009 U 51 50 8 269 - 227 126 154 0.10 45.8 9.7 2.8 0.7 3.2 154 6.3 5.2 Ca-Mg-HCO3 -1.96 -0.73 -2.87 -3.12 -9.59 2 -8.60 0.07
31136 U 20 50 5 333 - 270 144 186 0.11 54.7 12.0 3.4 1.1 4.4 176 10.0 8.9 Ca-Mg-HCO3 -1.79 -0.64 -2.62 -2.87 -9.39 2 -8.38 0.00
278 U 62 55 8 400 7.7 370 203 229 0.10 58.9 19.9 3.4 0.8 5.7 248 16.5 16.7 Ca-Mg-HCO3 0.14 0.21 -2.41 -2.66 -9.27 2 -7.14 0.04
Ab. Capuzzo U 68 56 5 441 7.8 380 150 238 0.08 64.7 18.6 2.8 1.2 2.9 183 101.9 5.0 Ca-Mg-HCO3-SO4 0.15 0.25 -1.63 -1.88 -9.65 2 -7.73 1.27
1374 U 95 60 6 512 7.9 431 136 288 0.11 83.3 19.4 4.2 0.8 4.1 165 149.0 4.6 Ca-Mg-SO4-HCO3 -0.32 0.05 -1.38 -1.62 -9.36 2 -8.66 0.33
31002 U 41 60 8 640 - 551 285 373 0.08 113.9 21.5 3.5 1.0 7.9 347 18.4 36.4 Ca-Mg-HCO3 -0.68 -0.07 -2.18 -2.43 -9.11 2 -6.43 0.04
31101 U 3 60 6 427 - 335 157 216 0.43 58.5 17.0 14.4 1.0 25.6 192 13.8 13.1 Ca-Mg-HCO3 -1.58 -0.59 -2.47 -2.73 -7.97 2 -7.73 0.00
614 U 21 61 8 552 7.4 504 249 313 0.14 79.0 28.1 5.7 1.0 10.9 304 40.5 35.7 Ca-Mg-HCO3 0.09 0.17 -1.97 -2.22 -8.77 2 -7.14 0.09
620 U 30 62 8 492 7.6 434 176 282 0.06 70.4 25.8 2.3 1.2 5.2 215 101.5 13.3 Ca-Mg-HCO3-SO4 -0.01 0.11 -1.61 -1.86 -9.48 2 -8.59 0.11
1520 U 76 62 10 528 7.6 486 265 305 0.04 80.2 25.4 1.7 0.4 6.4 324 13.7 34.2 Ca-Mg-HCO3 0.21 0.26 -2.42 -2.67 -9.53 - - -
608 U 24 63 8 618 7.3 580 301 356 0.13 92.7 30.2 5.7 1.7 11.4 367 34.2 36.9 Ca-Mg-HCO3 0.07 0.18 -2.00 -2.25 -8.77 2 -7.19 0.13
31067 U 64 65 7 407 - 336 175 229 0.13 66.4 15.3 4.7 2.0 5.7 213 13.2 15.9 Ca-Mg-HCO3 -1.44 -0.49 -2.47 -2.72 -9.16 2 -7.55 0.24
619 U 20 67 8 613 7.4 565 271 349 0.17 88.6 31.0 7.4 1.6 12.9 331 53.9 38.8 Ca-Mg-HCO3 0.26 0.25 -1.82 -2.07 -8.59 2 -7.31 0.08
237 U 95 70 12 514 7.5 467 211 286 0.18 78.2 22.0 6.8 4.4 9.3 258 69.2 16.2 Ca-Mg-HCO3-SO4 0.00 0.19 -1.74 -1.99 -8.81 2 -7.65 0.18
1365 U 87 70 6 504 7.6 487 262 298 0.04 73.1 28.0 1.6 0.3 8.5 319 13.4 43.6 Ca-Mg-HCO3 0.04 0.13 -2.47 -2.72 -9.43 - - -
31004 U 43 70 4 494 - 424 223 293 0.08 85.2 19.4 3.0 1.0 6.0 272 11.2 28.0 Ca-Mg-HCO3 -0.98 -0.27 -2.49 -2.74 -9.30 2 -8.47 0.86
250 U 73 72 13 538 7.4 497 255 308 0.19 80.1 26.2 7.8 0.9 13.0 310 36.0 19.4 Ca-Mg-HCO3 0.07 0.18 -2.00 -2.25 -8.60 2 -7.18 0.02
27 U 129 75 6 478 7.5 464 257 273 0.16 83.3 15.9 6.5 3.3 8.6 313 11.1 22.0 Ca-Mg-HCO3 0.09 0.31 -2.50 -2.75 -9.19 - - -
235 U 123 80 17 337 7.7 318 182 187 0.10 61.4 8.3 3.2 1.6 3.1 222 8.8 6.0 Ca-HCO3 -0.23 0.22 -2.63 -2.88 -9.55 2 -7.43 0.11
746 U 39 84 5 460 7.7 429 234 265 0.08 70.9 21.4 3.1 0.9 6.1 285 17.2 24.1 Ca-Mg-HCO3 0.33 0.33 -2.35 -2.60 -9.28 - - -
233 U 139 86 13 363 7.7 341 192 208 0.07 54.3 17.6 2.4 0.7 4.0 234 12.9 9.5 Ca-Mg-HCO3 0.07 0.18 -2.54 -2.79 -9.59 - - -
35 U 125 90 8 591 7.4 544 276 345 0.09 86.9 31.1 3.8 2.3 10.3 337 33.7 39.3 Ca-Mg-HCO3 0.18 0.21 -2.02 -2.27 -8.98 2 -7.06 0.03
38 U 118 90 8 577 7.4 529 261 339 0.10 86.5 29.9 4.4 1.8 10.2 319 51.6 26.5 Ca-Mg-HCO3 0.04 0.16 -1.84 -2.09 -8.93 2 -7.43 0.22
169 U 101 90 9 487 7.7 455 254 290 0.05 75.3 24.6 2.0 0.3 4.9 309 13.2 25.5 Ca-Mg-HCO3 0.31 0.30 -2.45 -2.70 -9.58 2 -7.02 0.12
Forgiarini U 140 90 2 477 8.0 401 215 270 0.04 65.6 25.9 1.4 0.4 4.5 263 18.9 22.1 Ca-Mg-HCO3 0.82 0.53 -2.33 -2.59 -9.75 - - -
240 U 76 91 11 498 7.5 459 235 291 0.09 73.4 26.2 3.5 1.2 7.5 286 35.4 25.2 Ca-Mg-HCO3 0.01 0.14 -2.05 -2.30 -9.15 2 -7.22 0.00
1528 U 124 91 9 548 7.6 531 279 327 0.12 83.9 28.4 4.8 0.2 7.0 341 17.2 48.4 Ca-Mg-HCO3 0.18 0.23 -2.33 -2.58 -9.07 2 -7.58 0.31
L U 35 96 1 351 7.9 336 187 208 0.03 51.1 19.4 1.1 0.2 5.3 228 7.7 22.9 Ca-Mg-HCO3 0.74 0.44 -2.80 -3.04 -9.79 - - -
Vacile Acq. U 153 104 4 329 7.9 300 168 183 0.04 55.0 11.2 1.3 0.5 2.2 205 16.4 8.0 Ca-Mg-HCO3 0.13 0.34 -2.43 -2.68 -10.08 - - -
605 U 134 120 9 351 7.9 326 186 204 0.04 52.2 17.9 1.2 0.8 3.1 227 10.0 13.9 Ca-Mg-HCO3 0.57 0.41 -2.67 -2.91 -9.99 2 -7.33 0.03
230 U 142 126 10 543 7.5 493 231 319 0.07 82.4 27.5 3.0 0.8 7.8 282 61.0 28.9 Ca-Mg-HCO3 0.02 0.16 -1.78 -2.03 -9.20 2 -7.63 0.10
177 U 118 132 8 423 7.8 389 214 242 0.05 62.3 21.0 1.7 0.7 4.9 261 9.4 27.6 Ca-Mg-HCO3 0.28 0.29 -2.64 -2.90 -9.63 2 -7.95 0.12
Table 1. Continued
δ18O Std.
Borehole ID Type Elevation Depth nch EC pH TDS Alk TH SAR Ca2+ Mg2+ Na+ K+ Cl- HCO3 - SO42- NO3- Water Type SIdol SIcal SIgyp SIani SIhal nis δ18O
dev.
(‰VSMO
(m a.s.l.) (m) (µS/cm) (mg/L) (mg/L) (mg/L) (meq/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)
W)
225 U 197 135 17 524 7.7 477 190 305 0.11 84.3 22.8 4.3 0.7 8.8 232 110.8 8.4 Ca-Mg-HCO3-SO4 0.24 0.32 -1.50 -1.75 -9.02 2 -8.38 0.07
1518 U 150 136 8 482 7.7 446 246 276 0.04 69.5 24.8 1.4 0.4 6.5 300 11.1 31.5 Ca-Mg-HCO3 0.10 0.19 -2.54 -2.80 -9.59 2 -7.17 0.19
601 U 154 150 8 427 7.7 395 224 248 0.03 66.1 20.1 1.0 0.3 3.9 274 9.1 20.4 Ca-Mg-HCO3 -0.07 0.14 -2.63 -2.88 -9.96 2 -7.59 0.06
171 U 224 176 9 575 7.7 534 280 339 0.05 81.2 33.1 1.9 0.4 9.1 341 16.7 50.7 Ca-Mg-HCO3 0.46 0.33 -2.35 -2.60 -9.34 1 -6.80 -
1315 SC 10 20 14 396 7.8 339 132 229 0.05 52.5 23.8 2.1 0.6 1.5 161 93.8 1.9 Ca-Mg-HCO3-SO4 0.07 0.11 -1.72 -1.97 -10.05 2 -9.09 0.12
1319 SC 11 20 13 396 7.8 341 129 230 0.05 52.9 23.8 2.0 0.6 1.4 158 98.4 1.8 Ca-Mg-HCO3-SO4 0.10 0.13 -1.70 -1.95 -10.10 2 -9.08 0.06
1516 SC 48 20 10 643 7.6 607 327 366 0.16 117.2 17.9 7.1 3.1 10.1 398 30.4 23.3 Ca-HCO3 0.39 0.50 -1.95 -2.20 -8.71 2 -7.07 0.23
1354 SC 32 22 9 367 7.9 338 197 211 0.03 53.5 18.8 1.1 0.4 3.4 240 6.7 14.6 Ca-Mg-HCO3 0.17 0.20 -2.83 -3.08 -9.97 2 -8.19 0.05
1525 SC 30 24 9 365 8.0 338 196 211 0.04 52.9 19.2 1.3 0.3 3.6 238 7.7 14.5 Ca-Mg-HCO3 0.21 0.24 -2.77 -3.02 -9.87 2 -8.07 0.08
1358 SC 24 25 9 418 7.8 389 221 241 0.04 60.6 21.9 1.6 0.5 5.1 269 8.3 21.3 Ca-Mg-HCO3 0.21 0.23 -2.71 -2.96 -9.64 2 -8.06 0.01
Ab. Zanin SC 29 25 4 391 7.8 350 203 216 0.03 53.9 19.8 1.1 0.3 3.7 248 8.3 15.3 Ca-Mg-HCO3 0.41 0.32 -2.74 -2.99 -9.95 2 -7.90 0.06
1301 SC 5 30 11 418 7.5 388 220 239 0.09 65.3 18.3 3.6 1.0 5.7 269 12.0 14.1 Ca-Mg-HCO3 -0.02 0.16 -2.51 -2.76 -9.24 2 -6.99 0.01
cim_s_vito SC 34 30 3 457 7.9 383 134 259 0.05 63.1 24.6 1.8 0.4 2.1 163 125.0 3.4 Ca-Mg-HCO3-SO4 0.20 0.21 -1.55 -1.80 -9.98 2 -8.82 0.48
1302 SC 2 31 15 389 7.6 366 208 222 0.06 58.7 18.4 2.8 0.9 4.5 254 10.7 14.6 Ca-Mg-HCO3 0.02 0.17 -2.59 -2.84 -9.46 2 -7.15 0.06
torrate_35 SC 18 35 6 471 7.9 408 157 270 0.06 69.5 23.4 2.3 0.5 3.1 192 110.6 6.2 Ca-Mg-HCO3-SO4 0.38 0.34 -1.56 -1.81 -9.70 - - -
1327 SC 6 37 14 499 7.5 472 264 288 0.09 82.3 20.0 3.9 1.1 7.1 322 15.4 18.7 Ca-Mg-HCO3 0.15 0.30 -2.34 -2.59 -9.12 2 -6.83 0.03
1300 SC 11 40 14 449 7.6 418 225 260 0.06 65.1 23.8 2.8 0.7 6.0 274 21.0 22.6 Ca-Mg-HCO3 0.19 0.22 -2.29 -2.54 -9.33 2 -7.19 0.08
1316 SC 12 40 10 417 7.8 364 142 240 0.07 56.6 24.0 2.5 0.6 2.2 173 102.3 3.3 Ca-Mg-HCO3-SO4 0.11 0.13 -1.67 -1.92 -9.85 2 -9.01 0.02
1320 SC 11 40 14 591 7.3 572 314 344 0.11 103.9 20.7 5.3 2.3 9.4 383 18.5 27.1 Ca-Mg-HCO3 0.08 0.29 -2.21 -2.46 -8.88 2 -6.85 0.01
1342 SC 44 40 9 479 7.8 415 156 275 0.07 74.1 22.0 2.6 0.6 3.1 191 114.7 7.0 Ca-Mg-HCO3-SO4 -0.19 0.08 -1.52 -1.77 -9.65 2 -8.59 0.25
1370 SC 23 40 3 470 7.8 404 146 271 0.07 69.7 23.6 2.5 0.5 2.9 178 121.3 5.4 Ca-Mg-HCO3-SO4 0.11 0.20 -1.52 -1.77 -9.69 2 -8.67 0.30
1373 SC 22 40 7 435 7.8 369 140 247 0.06 65.8 20.1 2.3 0.5 2.5 170 103.7 3.9 Ca-Mg-HCO3-SO4 0.17 0.24 -1.60 -1.85 -9.79 2 -8.41 0.12
167 SC 31 48 9 313 8.0 288 173 181 0.03 44.6 16.9 1.1 0.3 2.4 211 4.8 7.8 Ca-Mg-HCO3 0.14 0.21 -3.02 -3.27 -10.13 2 -8.44 0.01
Cimpello SC 17 50 2 306 7.8 289 167 180 0.07 41.4 18.7 2.3 0.5 1.5 204 17.8 2.6 Ca-Mg-HCO3 0.20 0.18 -2.49 -2.74 -10.00 - - -
1295 SC 17 54 14 591 7.5 562 297 346 0.08 94.5 26.7 4.1 1.1 9.6 362 32.4 30.4 Ca-Mg-HCO3 0.40 0.38 -2.01 -2.26 -8.98 2 -6.93 0.02
1367 SC 27 54 9 344 7.9 315 188 198 0.04 49.2 18.2 1.4 0.3 2.5 229 8.2 6.4 Ca-Mg-HCO3 -0.05 0.11 -2.75 -3.01 -10.01 2 -8.48 0.02
1335 SC 11 60 7 335 7.9 306 171 190 0.13 42.7 20.3 4.0 0.6 1.1 209 27.5 0.7 Ca-Mg-HCO3 0.36 0.23 -2.30 -2.55 -9.92 2 -8.48 0.39
1293 SC 19 70 14 451 7.5 428 237 260 0.07 76.6 16.8 3.0 1.2 5.5 289 12.6 21.0 Ca-Mg-HCO3 0.00 0.24 -2.44 -2.69 -9.33 2 -6.99 0.03
cim_zoppola SC 35 70 4 461 7.9 400 174 261 0.06 66.1 23.4 2.2 0.4 3.5 212 82.4 9.5 Ca-Mg-HCO3-SO4 0.57 0.43 -1.70 -1.95 -9.66 2 -8.26 0.03
Scuola
SC 15 70 5 413 7.9 353 134 235 0.06 57.6 22.2 2.3 0.5 2.0 164 102.2 2.6 Ca-Mg-HCO3-SO4 0.23 0.22 -1.66 -1.91 -9.91 - - -
media_70
180 SC 40 80 9 426 7.8 372 161 247 0.04 60.0 23.7 1.4 0.4 3.2 196 79.5 7.8 Ca-Mg-HCO3-SO4 -0.11 0.06 -1.74 -1.99 -9.90 2 -8.57 0.01
1296 SC 2 80 14 415 7.7 385 215 239 0.08 57.0 23.3 3.2 0.7 4.4 262 17.7 15.6 Ca-Mg-HCO3 0.32 0.25 -2.41 -2.66 -9.39 2 -7.43 0.13
1321 SC 2 80 14 514 7.5 482 250 297 0.14 69.7 29.9 6.3 0.8 10.1 305 35.4 24.1 Ca-Mg-HCO3 0.15 0.17 -2.06 -2.31 -8.75 - - -
1326 SC 1 80 14 513 7.5 473 246 294 0.15 68.7 29.7 6.9 0.9 12.6 300 32.4 20.6 Ca-Mg-HCO3 0.22 0.20 -2.10 -2.36 -8.62 2 -7.55 0.02
M SC 30 80 4 373 7.9 326 156 207 0.04 48.7 20.7 1.4 0.3 2.3 191 56.1 5.9 Ca-Mg-HCO3-SO4 0.19 0.20 -1.95 -2.20 -10.03 2 -8.51 0.08
Scuola
SC 15 80 3 400 8.1 346 132 232 0.07 52.0 24.9 2.4 0.5 1.5 161 101.2 2.1 Ca-Mg-HCO3-SO4 0.66 0.38 -1.70 -1.95 -9.98 2 -9.14 0.03
media_80
1378 SC 33 81 4 345 7.9 301 152 193 0.03 44.4 19.9 1.1 0.2 2.3 185 41.7 6.3 Ca-Mg-HCO3-SO4 0.11 0.15 -2.10 -2.35 -10.13 1 -8.57 -
1306 SC 5 90 14 361 7.8 331 183 193 0.30 44.4 19.9 11.1 1.0 1.0 224 29.1 1.4 Ca-Mg-HCO3 0.37 0.24 -2.27 -2.52 -9.52 2 -8.29 0.06
1307 SC 8 90 14 444 7.6 405 203 261 0.08 61.5 26.1 3.5 0.7 2.8 248 52.8 8.3 Ca-Mg-HCO3-SO4 0.04 0.11 -1.92 -2.17 -9.56 2 -8.27 0.04
Table 1. Continued
δ18O Std.
Borehole ID Type Elevation Depth nch EC pH TDS Alk TH SAR Ca2+ Mg2+ Na+ K+ Cl- HCO3 - SO42- NO3- Water Type SIdol SIcal SIgyp SIani SIhal nis δ18O
dev.
(‰VSMO
(m a.s.l.) (m) (µS/cm) (mg/L) (mg/L) (mg/L) (meq/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)
W)
1310 SC 9 90 14 372 7.8 330 158 211 0.13 47.2 22.6 5.2 0.8 1.0 193 58.6 0.7 Ca-Mg-HCO3-SO4 0.19 0.15 -1.95 -2.20 -9.84 2 -8.67 0.10
1312 SC 10 90 13 483 7.5 457 249 285 0.07 69.5 27.1 3.1 0.8 5.4 303 25.7 20.1 Ca-Mg-HCO3 0.13 0.17 -2.19 -2.44 -9.35 2 -7.60 0.04
1325 SC 3 90 11 512 7.6 481 248 293 0.15 71.6 27.7 7.1 0.9 13.1 302 31.5 25.3 Ca-Mg-HCO3 0.47 0.32 -2.11 -2.35 -8.60 - - -
1361 SC 37 91 4 387 7.9 381 216 234 0.03 57.8 21.7 1.2 0.2 4.6 263 7.6 18.3 Ca-Mg-HCO3 0.71 0.47 -2.76 -3.01 -9.83 1 -8.04 -
339 SC 3 100 10 410 7.7 380 216 232 0.20 51.1 25.4 7.0 1.3 3.7 264 26.8 0.2 Ca-Mg-HCO3 0.44 0.25 -2.28 -2.52 -9.16 2 -7.74 0.05
1309 SC 3 100 14 372 7.7 340 173 204 0.22 45.6 21.8 8.6 1.0 0.9 211 51.9 0.2 Ca-Mg-HCO3-SO4 0.26 0.16 -2.03 -2.27 -9.66 2 -8.62 0.02
1304 SC 20 110 14 540 7.5 507 258 313 0.14 79.2 28.0 6.9 1.2 10.6 314 36.0 30.2 Ca-Mg-HCO3 0.38 0.30 -2.02 -2.27 -8.71 2 -7.28 0.06
1328 SC 28 120 10 472 7.5 431 205 279 0.08 69.3 25.8 2.9 0.8 3.9 251 65.3 12.3 Ca-Mg-HCO3-SO4 -0.06 0.09 -1.79 -2.04 -9.49 2 -8.23 0.02
338 DC 0 120 14 371 7.9 364 226 139 1.37 30.9 14.9 37.0 1.7 1.1 275 2.0 0.1 Na-Ca-Mg-HCO3 0.51 0.28 -3.75 -3.99 -8.96 2 -9.59 0.06
1297 DC 3 120 13 376 7.8 350 208 214 0.16 47.6 23.1 6.0 1.2 2.7 254 14.3 0.1 Ca-Mg-HCO3 0.39 0.25 -2.55 -2.80 -9.34 2 -7.50 0.06
1324 DC 3 120 10 469 7.5 435 241 276 0.08 67.6 26.1 2.9 0.7 6.3 294 13.9 23.6 Ca-Mg-HCO3 0.43 0.28 -2.48 -2.72 -9.30 2 -7.36 0.28
1299 DC 2 144 14 445 7.6 416 228 260 0.07 63.7 24.5 3.1 0.8 5.9 278 17.8 19.7 Ca-Mg-HCO3 0.21 0.20 -2.39 -2.64 -9.30 2 -7.01 0.04
1364 DC 14 150 4 373 7.7 359 225 177 0.62 44.1 16.2 19.0 1.4 0.8 274 0.2 0.1 Ca-Mg-HCO3 0.19 0.21 -4.57 -4.82 -9.39 2 -10.35 0.06
1317 DC 12 165 14 403 7.8 344 127 231 0.05 53.0 23.9 2.1 0.6 1.6 155 104.2 1.9 Ca-Mg-HCO3-SO4 0.05 0.11 -1.68 -1.93 -10.03 2 -9.13 0.06
1350 DC 20 173 7 328 8.0 289 145 188 0.04 43.3 19.5 1.2 0.4 1.6 177 42.0 3.4 Ca-Mg-HCO3-SO4 0.25 0.21 -2.11 -2.36 -10.24 2 -8.47 0.30
342 DC 20 174 14 401 7.9 347 127 233 0.06 53.1 24.4 2.0 0.5 1.7 155 105.9 1.5 Ca-Mg-HCO3-SO4 0.08 0.12 -1.67 -1.92 -10.02 2 -9.18 0.01
1294 DC 11 177 14 426 7.6 393 211 246 0.08 58.7 24.2 3.4 0.8 5.5 258 28.2 12.9 Ca-Mg-HCO3 0.14 0.17 -2.19 -2.45 -9.28 2 -7.65 0.10
1323 DC 9 180 14 366 7.8 319 137 209 0.08 46.5 22.6 3.0 0.7 1.1 167 74.1 1.3 Ca-Mg-HCO3-SO4 0.13 0.11 -1.86 -2.11 -10.02 2 -8.97 0.08
1345 DC 36 180 4 435 7.8 381 168 247 0.07 61.1 22.8 2.6 0.4 2.0 204 86.3 1.2 Ca-Mg-HCO3-SO4 0.14 0.20 -1.70 -1.95 -9.83 1 -8.69 -
Doncal DC 32 180 7 466 7.9 394 136 267 0.05 65.1 25.3 1.7 0.4 3.0 165 129.3 3.6 Ca-Mg-HCO3-SO4 0.19 0.22 -1.52 -1.77 -9.85 2 -8.95 0.01
casa_riposo_
DC 14 183 4 376 8.0 329 143 216 0.09 46.9 24.1 2.9 0.5 1.2 175 77.2 1.6 Ca-Mg-HCO3-SO4 0.49 0.29 -1.85 -2.10 -9.99 2 -8.91 0.06
183
1330 DC 21 190 14 389 7.8 343 134 228 0.05 53.8 22.8 1.9 0.5 2.4 163 93.9 2.1 Ca-Mg-HCO3-SO4 -0.08 0.06 -1.71 -1.96 -10.01 2 -9.07 0.04
torrate_190 DC 18 190 7 391 8.0 338 141 225 0.06 51.1 23.7 2.1 0.4 1.7 172 83.3 3.2 Ca-Mg-HCO3-SO4 0.39 0.27 -1.78 -2.03 -9.99 2 -8.92 0.00
1314 DC 2 200 12 369 7.7 329 165 200 0.21 44.8 21.4 8.2 0.8 1.0 201 51.4 1.4 Ca-Mg-HCO3-SO4 0.21 0.13 -2.04 -2.27 -9.66 - - -
1340 DC 17 200 7 421 7.8 414 255 143 1.57 35.9 12.9 43.0 4.6 1.2 311 0.5 0.3 Na-Ca-Mg-HCO3 0.47 0.33 -4.34 -4.59 -8.83 2 -10.32 0.02
1515 DC 10 200 7 340 7.9 330 208 173 0.49 39.7 17.9 14.8 1.2 0.9 253 0.7 0.8 Ca-Mg-HCO3 0.55 0.34 -4.05 -4.29 -9.41 2 -8.34 0.05
343 DC 12 220 9 385 7.7 330 141 214 0.16 47.8 23.0 5.5 0.6 1.4 172 77.7 2.7 Ca-Mg-HCO3-SO4 -0.05 0.02 -1.84 -2.08 -9.68 2 -8.92 0.05
torrate_283 DC 18 283 1 334 7.7 308 193 153 0.39 35.5 15.6 11.0 1.0 0.5 235 6.4 0.5 Ca-Mg-HCO3 -0.02 0.06 -2.98 -3.23 -9.80 - - -
U unconfined, SC shallow confined, DC deep confined; nch number of chemical analyses, nis number of isotopical analyses; SI saturation indices; subscript dol dolomite, cal calcite, gyp gypsum, ani anidrite, hal halite; VSMOW Vienna standard mean
ocean water; Std. dev. standard deviation
Table 2. Univariate overview of the groundwater samples. All values are in mg/l, unless otherwise indicated. Std. dev. = standard
deviation, n = number of samples.

Parameter Unconfined Shallow Confined Deep Confined


Min. Max. Mean Std. dev. n Min. Max. Mean Std. dev. n Min. Max. Mean Std. dev. n

Temp (°C) 11.6 15.7 13.5 0.9 86 11.7 17.7 14 1.4 43 12.3 19.3 14.7 1.9 20
EC (µS/cm) 260.3 935.6 509 114 86 306 643 432.7 74.4 43 327.9 469.2 393.3 40.4 20
pH 6.9 8 7.5 7.7 78 7.3 8.1 7.7 8 43 7.5 8 7.8 8.2 20
TDS 211.8 927.6 466.6 114.2 86 288.4 607.4 393.7 74.9 43 288.6 435.3 355.6 38.6 20
Ca 44.5 149.6 79.2 18.1 86 41.4 117.2 62.5 15.9 43 30.9 67.6 49.5 10.2 20
Mg 7.9 45.8 23.4 6.9 86 16.8 29.9 22.5 3.4 43 12.9 26.1 21.4 3.9 20
Na 1 14.4 4.1 2.3 86 1.1 11.1 3.5 2.3 43 1.2 43 8.7 11.8 20
K 0.2 4.4 1.1 0.7 86 0.2 3.1 0.8 0.5 43 0.4 4.6 1 0.9 20
Cl 2.1 25.6 7.7 3.9 86 0.9 13.1 4.4 3.2 43 0.5 6.3 2.2 1.7 20
HCO3 140.3 651.8 287.2 77.6 86 157.8 398.2 240.3 61.5 43 154.8 311.4 217 53.1 20
SO4 6.2 149 39.2 35.1 86 4.8 125 47.6 38.4 43 0.2 129.3 50.5 42.6 20
NO3 0.2 50.7 23.7 13.2 86 0.2 30.4 11.6 9.1 43 0.1 23.6 4.1 6.6 20
18
δ O (‰) -8.9 -6.25 -7.59 0.7 72 -9.14 -6.83 -8.05 0.7 38 -10.35 -7.01 -8.74 0.9 18
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 245

Geochemical Characteristics of the Groundwater


Sampling and Analytical Procedures

In order to have hydrochemical data covering the entire area of the Friuli Venezia Giulia
Plain 149 municipal and private wells have been studied for a total of 3249 analyses (Table
1). The water temperature, electrical conductivity (EC) and pH of each sample were measured
in situ by ARPA FVG with a conductivity meter standardised to 20 °C, and a pH electrode
previously calibrated with standard buffers. The water samples were chemically analysed by
ARPA FVG by atomic absorption spectrometry (Ca2+, Na+, K+, Mg2+, accuracy 5–10%) and
ion chromatography (Cl-, SO42-, NO3-, HCO3-, accuracy 5–10%). TDS, Alk, TH and SAR
were calculated using Aquachem version 4.0 whereas the saturation indices (Table 1) were
calculated with PHREEQC (Parkhurst et al., 1980). Stable isotope analyses of water
(including rainfall) were measured on a VG Optima mass spectrometer at the University of
Trieste. The values are reported as per mil deviations from the VSMOW standard using the
conventional δ notation (Craig, 1961). The standard deviation of the measurements is of about
±0.02‰ (2σ). The 87Sr/86Sr ratios were measured using a VG ISOMASS 54E mass
spectrometer with the 87Sr/86Sr ratios normalized to a 86Sr/88Sr ratio of 0.1194. During the
analysis the NBS 987 standard gave an average value of 0.71024 and no corrections were
applied to the measured Sr isotopic ratios.

Geochemistry

The dominant ion in the groundwater samples (Figure 8) from the unconfined and confined
aquifer is HCO3 and five hydrochemical facies have been found: Facies 1: Ca-HCO3; Facies
2: Ca-Mg-HCO3; Facies 3: Ca-Mg-HCO3-SO4; Facies 4: Ca-Mg-SO4-HCO3 and Facies 5:
Na-Ca-Mg-HCO3, Table 1). Facies 1 to facies 4 have been recorded in all the samples,
whereas Na-Ca-Mg-HCO3-type waters (Facies 5) are located only in confined aquifers
beneath clay and silt layers, likely reflecting exchange reactions by water-rock and microbial
interactions. Groundwater generally shows a conductivity and mineralisation decrease with
increasing depth (Table 2). The pH values range from 6.9 to 8.0 (mean of 7.5) to 7.3 to 8.1
(mean of 7.7) to 7.5 to 8.0 (mean of 7.8) in the unconfined, shallow and deep confined
groundwater respectively (Table 2). Mean concentrations of HCO3 and Mg are generally
similar between unconfined and confined groundwaters, whereas Na concentrations increase
with increasing depths for confined groundwater (from 3.5 to 8.7 mg/l). Potassium
concentrations are similar for unconfined and confined groundwaters. Chloride concentrations
range from 2.1 mg/l to 25.6 mg/l (mean of 7.7 mg/l) in the samples from the aquifer and from
0.9 mg/L to 13.1 mg/l (mean of 4.4 mg/l) in shallow confined groundwater and from 0.5 mg/l
to 6.3 mg/l (mean of 2.2 mg/l) in the deep confined aquifers. Nitrate concentrations range
from 0.2 mg/l to 51 mg/l (with one sample is above the limit of the Italian Law of 50 mg/l),
with a mean of 23.7 mg/l, in the unconfined groundwater and from 0.2 mg/l to 30.4 mg/l
(mean of 11.6 mg/l) in shallow confined groundwater and from 0.1 mg/l to 23.6 mg/l
(although only 15% of these groundwater samples have values of more then 4 mg/l), and a
mean of 4.1 mg/l, in the deep confined aquifers. Calcium concentrations range from 44.5 mg/l
to 149.6 mg/l (mean of 79.2 mg/l) in the unconfined samples and from 41.4 mg/l to 117.2
246 Franco Cucchi, Giuliana Franceschini and Luca Zini

mg/l (mean of 62.5 mg/l) in shallow confined groundwater and from 30.9 mg/l to 67.6 mg/l
(mean of 49.5 mg/l) in the deep confined aquifers. In contrast to the elements discussed
above, the distribution of SO4 is very different. Sulphate concentration decreases with the
distance to the Tagliamento River that has 130 mg/l of sulphate when its waters enter in the
Friuli Plain due to gypsum formations outcropping in the Carnic Alps.

Figure 8. SO4, Piper diagram of the unconfined and confined groundwater samples.

Environmental Isotopes

The δ18O of groundwater (Figure 9) fall in the range -6.25‰ to -8.90‰ with a mean value of
-7.59‰ in the unconfined aquifer, from -6.83‰ to -9.14‰ with a mean value of -8.05‰ in
the shallow confined aquifer and from -7.01‰ to -10.35‰ with a mean value of -8.74‰ in
the deeper confined aquifer (Tables 1 and 3). The most depleted groundwater samples are
located in the Carnic Alps (Table 1). The rainfall samples show a great variation in isotopic
composition (calculated over a two year monitoring period), the δ18O weighted mean values
range from -7.09‰ to -10.39‰ with an average value of -9.04‰ in the mountainous areas
and from -7.45‰ to -7.78‰ for δ18O in the Friuli Venezia Giulia Plain with a mean of -
7.62‰ (Table 3).
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 247

Figure 9. SO4, Ca/Mg ratio, NO3 and atrazine distribution for the layers A to E of the confined aquifers
(modified after Cucchi et al., 1998).

Table 3. Sampling period and δ18O minimum, maximum, standard deviation and mean
weighted values for the rainfall samples.

Mean Std.
Pluviometer Elevation Time perioda Total volume Minimum Maximum
weighted dev.
of values of of
ID Location From To values of δ18O values of δ18O 18
precipitation δ18O δ O
(m a.s.l.) (mL) (‰VSMOW) (‰VSMOW) (‰VSMOW)
P1 Tarvisio 794 30/09/04 - 14/09/06 93000 -20.75 -4.61 -10.39 3.9
P2 Saletto 496 01/09/04 - 14/09/06 122000 -16.91 -4.66 -8.60 3.2
P3 Musi 526 07/10/04 - 15/09/06 145550 -15.43 -2.77 -7.88 3.0
P4 Matajur 1326 07/10/04 - 16/09/06 119900 -13.66 -4.81 -8.83 2.4
P5 Gemona 184 28/09/04 - 13/09/06 125400 -12.04 -1.68 -7.09 2.6
P6 Enemonzo 438 28/09/04 - 13/09/06 95000 -19.70 -3.73 -8.95 4.0
P7 Zoncolan 1755 05/10/04 - 14/09/06 80300 -15.67 -5.20 -10.11 2.7
P8 Forni di Sopra 922 05/10/04 - 13/09/06 103900 -21.33 -4.72 -10.30 4.2
P9 Cansiglio 1033 11/09/04 - 13/09/06 121000 -19.54 -4.78 -9.33 2.9
P10 Roveredo 82 18/09/05 - 17/09/06 30200 -14.29 -4.96 -7.78 3.0
P11 Barcis 460 16/09/05 - 15/09/06 66800 -14.68 -4.53 -8.91 3.2
P12 Palmanova 29 16/09/05 - 18/09/06 23300 -12.63 -3.45 -7.45 2.4
P13 Paularo 633 15/09/05 - 14/09/06 44900 -17.79 -4.19 -9.06 4.0
VSMOW Vienna standard mean ocean water; Std. dev. standard deviation
a
two-years for P1 to P9 and one-year for P10 to P13

The most isotopically depleted rainfall samples were those collected during winter months
with, for example, a value of -21.33‰ recorded for Forni di Sopra in January-February 2005.
The vertical isotopic gradient is 0.29‰ per 100 m, in accordance with previous findings for
248 Franco Cucchi, Giuliana Franceschini and Luca Zini

Friuli Venezia Giulia (Longinelli and Selmo, 2003; Longinelli et al., 2006) and for
neighbouring Slovenia and Croatia (Vreča et al., 2006) as well as for Austria (Kralik et al.,
2006). Stable isotopes in the Friuli Venezia Giulia river waters show different values
according to the mean altitude of their drainage basin. However, δ18O river water samples
collected from 2005 to 2006 range from -8.23‰ to -9.31‰ for the Tagliamento River, from -
7.68‰ to -9.10‰ for the Isonzo River, from -7.63‰ to -8.60‰ for the Natisone River, from
-7.94‰ to -9.78‰ for the Cellina River and from -7.09‰ to -9.98‰ for the Livenza River
(CAMI, 2007).
Several observations may be made about the stable oxygen isotope composition of the
groundwater in relation to the surface waters as well as to modern rainfall. Firstly, the
groundwater samples from the unconfined aquifer have values similar to a mixture between
precipitation and local river infiltrations. Secondly, the groundwaters, in the shallow confined
aquifers have depleted values, up to 1.52‰, relative to modern Plain rainfall (Tables 1 and 3).
Thirdly, the deep confined groundwater is notably depleted in comparison to both Plain
rainfall (up to 2.73‰) and the shallower groundwater (up to 1.21‰).

Carbon Isotope Systematics

Carbon system parameters and δ 13C values of dissolved inorganic carbon (DIC) are important
in determining the sources of carbon and tracing water–rock and microbial interactions. The
14
C data shown in Table 4 support many of the points made above and allow further
constraining the flow system and recharging periods. Percent modern carbon (pmc) values
range from 36 to 48 in the shallow confined groundwater to 0.8 to 0.9 in the deep confined
groundwater indicating a good correlation between pmc and depth (Table 4). The presence of
organic source of dead carbon affects the 14C activity correlation, thus precluding precise
dating. However, residence times have been calculated using the model described by Ingerson
and Pearson (1964) which is the most suitable model for terrigenous aquifers and has been
applied in similar groundwater conditions in the neighboring region of Veneto as well as in
Lombardy (Pilla et al., 2006; Zuppi and Sacchi 2004). The following values have been used
in age calculation: the activity of soil CO2, which is equal to 100 pmc, -25‰ as δ 13C; the
activity of inorganic carbonates within the aquifer matrix, which are equal to 0 pmc and
δ13C=0‰. Using this calculation the variation in pmc was interpreted as reflecting differences
in groundwater residence time rather than relative dissolution with old radioactively dead
radiocarbon. Several shallow groundwater samples (Table 5) have negative ages that almost
certainly represent groundwater recharged since atmospheric nuclear testing in the 1960s,
which produced atmospheric CO2 with pmc contents of up to 200. The shallow confined
groundwater residence time is ~100 years, whereas the deep confined groundwater is ~23 to
27 ka (Figure 10). The occurrence of relatively old groundwater in the deep confined waters
is consistent with its high clay contents as a consequence of its low hydraulic conductivities
(Mosetti, 1983). Accordingly to the radiocarbon data presented in this work, the shallow
groundwater moves at a speed of 1 km/yr, whereas the deep confined groundwater only 3
m/yr (Figure 10).
Table 4. Chemical, isotopical and radiocarbon data for four selected samples of the confined and unconfined groundwater samples.

Borehole
Type Elevation Depth T δ18O dD d C13 C14 EC pH TDS Alk TH SAR Ca2+ Mg2+ Na+ K+ HCO3 - SO42- NO3- Cl-
ID
(m a.s.l.) (m) °C ‰ ‰ % (µS/cm) (mg/l) (mg/l) (mg/l) (meq/l) (mg/l) (mg/l) (mg/l) (mg/l) (mg/l) (mg/l) (mg/l) (mg/l)
339 SC 3 100 17.5 -7.63 -50.55 10.49 -9.26 36.70 409.9 7.7 379.6 216.4 231.9 0.20 51.1 25.4 7.0 1.3 263.9 26.8 0.2 3.7
1317 SC 12 165 13.4 -9.08 -61.68 10.96 -7.37 47.50 402.6 7.8 344.3 126.9 230.7 0.05 53.0 23.9 2.1 0.6 154.8 104.2 1.9 1.6
1340 DC 17 200 16.9 -10.33 -72.31 10.33 -3.45 0.80 421.3 7.8 413.5 255.4 142.9 1.57 35.9 12.9 43.0 4.6 311.4 0.5 0.3 1.2
1364 DC 14 150 14.3 -10.30 -72.17 10.23 -6.39 0.90 372.8 7.7 358.6 224.7 176.7 0.62 44.1 16.2 19.0 1.4 274.0 0.2 0.1 0.8
Scuola SC 15 80 15.1 -9.12 -62.13 10.83 -8.69 48.30 400.3 8.1 345.6 132.0 232.2 0.07 52.0 24.9 2.4 0.5 161.0 101.2 2.1 1.5

Figure 10. A schematic reconstruction of groundwater recharges and movement for the Friuli Venezia Giulia Plain, in which the shallower confined
groundwater moves up to 1 km/year whereas the deep confined groundwater only approximately 3 m/yr.
250 Franco Cucchi, Giuliana Franceschini and Luca Zini

Geochemical Data Interpretation


Cation exchange between the deep confined groundwater and the silty-clay layers are
responsible for the groundwater chemistry variations with depth from Ca-Mg-HCO3, Ca-Mg-
HCO3-SO4, Ca- HCO3 facies to Ca-Mg-HCO3, Ca-Mg-HCO3-SO4, Na-Ca-Mg- HCO3 facies.
The mean saturation indices for calcite and dolomite were calculated to be from 0.12 to 0.23
(except for the unconfined groundwater samples where they show marked variations from -
0.16 to 0.53) therefore the aquifers are at or near equilibrium with respect to these two
minerals; whereas all of the water samples are undersaturated in gypsum, anhydrite and halite
(Table 1).
The large variations between isotopic signature as well as the different chemical signature
between the unconfined and shallow confined waters in comparison to the deep confined
groundwater indicate that their hydraulic conditions have significantly changed. In addition to
the stable isotope data, the radiocarbon values indicate that there is very little continuity
between the aquifers and that the deeper aquifers are not directly recharged by present rainfall
and river infiltrations. Significant late Quaternary sea-level fluctuations, associated with
alternating cooler and wetter periods, have changed the hydraulic gradients and partially or
completely disconnected the deeper parts of the confined aquifer systems from the more
active surface circulations. Similar conditions have been described for other part of the Po
Plain, where during colder conditions of the LGM alpine glaciers reached the present day
plain and the sea-level lowstand (of approximately -120 m) resulted in a more depleted
groundwater isotopic fingerprint (Zuppi and Sacchi, 2004; Pilli, 2005; Pilla et al., 2006).

Paradiso Wells: A Case Study

An explanation of the different chemical and stable isotope characteristics that have allowed
separating the shallow confined (layers A and B) from the deep confined groundwaters are
well shown in the Paradiso wells. In fact, in the Paradiso farm four different wells are tapped
for bottled waters. The Paradiso groundwater is extracted from the shallow confined aquifer
at depth of 17 to 25 m. The Annia and Torsa groundwaters are pumped from the shallow
confined aquifer at depth of 82 to 96 m and 100 to 126 m, respectively. The Pocenia bottled
groundwater is extracted from the deep confined groundwater at depth of 200 to 213 m.

Table 5. Uncorrected and corrected ages of the four confined and unconfined
groundwater samples.

Residence Time Residence time


Sample Uncorrected A0=100 Corrected Pearson model
1340 39900 23542
1364 38900 27663
scuola 6000 R
1317 6450 R
339 8300 76
Table 6. Chemical and isotopical composition of groundwater samples from the Pocenia wells.

T EC pH Na+ K+ Ca2+ Mg2+ NH4+ Fe Cl- NO3- SO42-HCO3- Sr 87


Sr/86Sr δ18O δ2 H δ13C A 14C Tritium Residence time Residence time
(T.U.)
μs/cm mg/l mg/l mg/l mg/l mg/l ppb mg/l mg/l mg/l mg/l mg/l ‰V- ‰V-SMOW V-PDB uncorrected corrected
SMOW A0=100 Pearson

PARADISO 13,7 500 7,7 3,4 0,7 74,8 28,5 <0,050<0.01 4,8 17,7 62,8 271 0,381 0.707974 -8,15 -55,25 -10,5 65.3±1 10,83 3500 R
(20)
ANNIA 14,2 407 8,0 3,4 0,6 54,8 24,6 <0,050<0,01 1,3 2,7 77,1 186 0,761 0.707823 -8,74 -59,22 -8,98 43.4±0.8 10,27 6900 R
(24)
TORSA 15,0 390 8,0 3,4 0,6 51,5 24,5 <0,050<0,01 1,0 2,0 81,0 171 0,659 0.707863 -8,88 -59,95 -8,94 39.8±0.7 8,74 7600 R
(35)
POCENIA 16,6 376 8,0 7,3 0,7 46,0 22,8 0,140 0,07 1,0 <0,1 68,4 183 0,734 0,70779 -8,84 -60,01 -9,18 14±0.5 8,4 16250 7971
252 Franco Cucchi, Giuliana Franceschini and Luca Zini

Groundwater samples were collected from outside taps in November 2007. On these samples,
physical and geochemical analyses have been performed to underline the different
characteristics of these waters (Table 6). The samples show a downward increase in water
temperature (from 13,7 to 16,6°C), NH4 concentration (from below detection limit value of
0.050 to 0.140 mg/l in the Pocenia well), Fe concentration (from below detection limit value
of 0.01 to 0.07 mg/l) as well as for the Sr content (from 0.381 mg/l to 0.734 mg/l). Marked
decreases have been recorded for EC (from 500 to 376 mg/l), Ca (from 74.8 to 46.0 mg/l),
Mg (from 28.5 to 22.8 mg/l), Cl (from 4.8 to 1.0 mg/l) and NO3 (from 17.7 mg/l to below the
detection limit of 0.1 mg/l). Depletion of the environmental stable isotope of oxygen (from -
8.15 to -8.84‰) and hydrogen (from -55.25 to -60‰) has also been found. 14C activities
range from 65.3 to 14 pmc and from -10.5 to 9.18‰ for δ13C (Table 6). The presence of an
organic source of dead carbon affects the 14C activities like any other alluvial aquifers that
preclude precise dating. Nevertheless, residence times have been calculated using the model,
first described by Ingerson and Pearson (1964), which is the still the most suitable model for
terrigenous aquifers and that have been previously used with similar purposed in the Padain
Plain (Pilla et al., 2007; Zuppi and Sacchi 2004). Groundwater corrected age indicates that
recharge for the shallow confined and deep confined aquifers took place in different periods:
the present for the shallow confined aquifer (up to 126 m), while for the palaeogroundwaters
of the deep confined aquifer (from 200 to 215 m) recharge occurred in the Holocene around 8
ka BP in the late Wurm Pluvial (MIS 2). The recharge periods of late Wurm for deep
confined groundwaters have been also found in other part of the Padain Plain, suggesting that
this probably was the last pluvial periods that have predominantly recharged these aquifers.
This information, coupled with the high quality standards of deep confined groundwaters (in
particular regarding nitrates), raises concern about their free exploitation.

Groundwater Provinces
Combined evidence from major ion, physical characteristics and stable isotope data suggest
that the overall geochemistry of the Friuli Venezia Giulia Plain alluvial aquifers can be
summarized into four different hydro-geochemical provinces for the sub-surface (unconfined
and shallow confined, Zone 1s to 4s of Figure 11) and deep confined groundwaters (Zone 1d
to 4d of Figure 12).

Zone 1 (Cellina-Meduna Province) the sub-surface groundwaters (Zone 1s, Figure 11)
are characterised by depleted δ18O values, high bicarbonate (217 to 460 mg/l) and nitrate (due
to intensive agricultural activity) ion (17 to 43 mg/l) and very low sulphates (0 to 50 mg/l)
and potassium (0.2 to 1.1 mg/l) concentrations. The deep confined groundwaters (Zone 1d,
Figure 12) have very high sodium (from 9.1 to 43 mg/l), high potassium (more then 1.8 mg/l)
and magnesium (more then 16 mg/l), low sulphate, nitrate and chlorides concentrations. This
province has the most depleted δ18O values (from -9 to 10.4‰, Figure 9) and all the samples
are subsaturated with regards to gypsum.
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 253

Figure 11. Sub-surface groundwater provinces (Zone 1s to 4 s).

Zone 2 (Tagliamento River Province) the subsurface groundwaters (Zone 2s, Figure 11)
are characterised by very high concentrations of sulphates (up to 125 mg/l) close to the
Tagliamento River and chlorides (6.4 to 11.9 mg/l) in the east, due to intensive human
activities. This zone has low bicarbonate (140 to 217 mg/l) and nitrate ion concentrations (0
to 17 mg/l). The deep confined groundwaters (Zone 2d, Figure 12) have high sulphate and
magnesium concentrations, whereas the concentrations of chlorides, bicarbonates, nitrate and
sodium are low.

Zone 3 (the Province between the Torre and Natisone rivers) the sub-surface
groundwaters (Zone 3s, Figure 11) have by high concentrations of nitrates (17 to 43 mg/l) due
to intense agricultural activity and high calcium (60 to 100 mg/l) and magnesium values (up
to 32 mg/l) due to local rivers, which drain carbonate-rich mountainous areas. This province
has δ18O values ranging from -6.5 to -7‰ indicative of good rainfall infiltration rates. The
deep confined groundwaters (Zone 3d, Figure 12) have high magnesium and low
bicarbonates, nitrates, chlorides, potassium and calcium indicative of low anthropomorphic
impact.
254 Franco Cucchi, Giuliana Franceschini and Luca Zini

Figure 12. Deep confined groundwater provinces (Zone 1d to 4d).

Zone 4 (Isonzo River Province) the sub-surface groundwaters (Zone 4s, Figure 11) have
high chlorides (6.4 to 11.9 mg/l) concentrations and low bicarbonates, sulphates, nitrates and
magnesium and δ18O values and are sub-saturated with regard to dolomite. The deep confined
groundwaters (Zone 4s, Figure 12) have characteristics similar to the sub-surface
groundwater, expect for higher magnesium values, up to 26 mg/l.

Conclusion
Groundwater in the Friuli Plain is commonly used for stock watering, irrigation and
increasingly for domestic supply. Escalating population and economic activities in the region
are intensifying groundwater usage. The combination of major ions, stable isotopes and
radiocarbon dating has enabled several aspects of the hydrogeology and hydrochemistry of
the sampled aquifer systems to be constrained. Results from this study indicate that the
isotopic content of unconfined and shallow confined groundwaters display fast recharge that
makes these groundwaters susceptible to contamination by pesticides and fertilizers as well as
by discharge from industries and urban areas. The unconfined and shallow confined
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 255

groundwaters have been subdivided into four different hydrological zones (Cellina-Meduna
Plain, Tagliamento River Plain, Plain between Torre and Natisone rivers and the Isonzo River
Plain) based on chemical ion distributions and geohydrological factors. A distinct and
complex hydrogeological circulation occurs within the deep confined groundwater. Lower
isotopic values of these groundwaters, compared to rainfall water, as well as lower
radiocarbon content show that these waters are not being recharged by local rainfall in a
relatively short time-scale. The deep confined waters in the Friuli Plain have passive
hydrodynamic conditions, with almost no cross-transfer with polluted shallower
groundwaters. To date, agricultural contaminations such as nitrate, have not impacted the
deeper groundwater. Since groundwater contributes to surface water, over pumping and or
contamination of shallow and deep aquifers may results in increased risk to water resources
throughout the entire region catchments. The integration of all stratigraphic data in a GIS
database as well as radiocarbon dating on selected deep confined groundwater are presently
underway and will provide additional and useful information for groundwater recharge
periods, exploitation and rock-water interactions making this research vital for water
management strategies.

Acknowledgements

The authors thank the Stable Isotope Laboratory of the University of Trieste for stable oxygen
analyses, Alberto Carniel, Enio De Corte, Anna Lutman, Giorgio Matassi and Manlio Princi,
of Arpa FVG for the geochemical data of groundwater samples. In addition, we thank the
GGACI working group, in particular Ivonne Burla, Susanna Erti, Alessio Mereu and
Francesco Treu, for sample collection and data management.

References
Basin Authority for the North Adriatic Rivers (1997) Piano di bacino del fiume Tagliamento-
Piano Stralcio per la sicurezza idraulica del medio e basso corso. [Autorita` di Bacino
Nazionale dei Fiumi Isonzo, Tagliamento, Livenza, Piave, Brenta-Bacchiglione]. Venice,
237 pp.
Bosellini, A (2004) The western passive margin of Adria and its carbonate platforms. The
Geology of Italy, Crescenti, U., D'Offizi, S., Merlini, S. & Sacchi, R. (eds) , 79–92.
CAMI (2007) Water-bearing characterization with integrated methods. http://www.cami-
life.net/eng.lived, cited 27 June 2007.
Carulli GB (2006) Carta geologica del Friuli Venezia Giulia (scale 1:150000) [Geological
Map of Friuli Venezia Giulia, scale 1:150000]. Ed. S.E.L.C.A. Firenze.
Cati A, Fichera R, Capelli V (1987) Itaia nordorilentale. Interpretazione integrata dei dati
geofisici e geologici [Geophysical and geological data interpretation of north-western
Italy]. AGIP San Donato Milanese, 121 pp.
Craig H (1961) Isotopic variations in meteoric waters. Science 133:1702-1703
Cucchi F, Franceschini G, Zini L, Treu F (2007) Hydrogeochemical and oxygen isotope
investigations in multilayer aquifers of the Friuli Venezia Giulia Plain, north-eastern
256 Franco Cucchi, Giuliana Franceschini and Luca Zini

Italy.Special publication of the XXXV Congress of the International Association of


Hydrogeologists (IAH) Lisbon 2007.
Cucchi F, Giorgetti F, Gemiti F, Massari G, Oberti S (1998) Caratterizzazione geochimica
delle falde acquifere della pianura friulana [Geochemical caractherisation of the Friulian
Plain aquifers]. Atti Acque sotterranee: risorsa invisibile 44:61-71.
Cucchi F, Massari G, Oberti S (1999) Il chimismo delle falde freatiche e artesiane della
Pianura Friulana [Chemistry of the phreatic and artesian aquifers in the Friulian Plain].
Quaderni Museo Carsico Geologico e Paleontologia 5:3-20.
Cucchi F, Piano C, Marinetti E, Massari G, Oberti S, Zini L (2000) Studies for the realization
of the hydrogeological map of Friuli Venezia Giulia. Ipogea 3:57-71.
Fontana A (2006) Evoluzione geomorfologica della bassa pianura friulana
[Geomorphological evolution of the lower Friulian Plain]. Ed Museo Friulano di Storia
Naturale Udine; 46 pp.
Fontana A, Bondesan A (2006) Il Tagliamento nella bassa pianura, tra dossi e incisioni
fluviali [The Tagliamento River in the lower Friulian Plain]. Ed. Circolo Menocchio
16:127-145
Foramitti R (1990) Il Tagliamento: studi e progetti, un fiume che deve unire [The
Tagliamento River: studies and projects]. Ed. Altan 20:133-88.
Granati C, Martelli G and Roda C (2000) “Valutazione preliminare del volume di acqua
estratta annualmente in Provincia di Udine dal sottosuolo della Bassa Pianura Friulana”.
IGEA, 15: 13-26.
Granati C., Martelli G. and Roda C (2001) Characterisation of the Artesian Aquifers in the
Friuli Plain (NE Italy) by Means of Low-Quality Hydrological Data. Journal of
Conference - EUG XI – Strasbourg – 8/12 April 2001: 48.
Martelli G and Granati C (2007) Valutazione della ricarica del sistema acquifero della bassa
pianura friulana [Recharge data from the Lower Friulian Plain]. Giornale di Geologia
Applicata 5, 89-114.
Martelli G, Granati C, Toscani L, Iacumin P, Selmo E (2007) Risultati preliminari delle
indagini isotopiche svolte sulle acque delle falde profonde della Bassa Pianura Friulana
[Prelimian results from istopical investigation of the deep aquifers of the Lower Friulian
Plain]. Giornale di Geologia Applicata 6, 93-101.
Ingerson, E and Pearson FJ (1964) Estimation of age and rate of motion of groundwater by
the 14C-method. In: Recent Researches in the Fields of Atmosphere, Hydrosphere, and
Nuclear Geochemistry.
Kralik M, Papesch W, Sticheler W (2006) Austrian network of isotopes in precipitation
(ANIP) and the good status in Alpine Karst water. Abstract presented at the 3rd KATER
International Conference, Vienna, October 2006.
Longinelli A, Anglesio E, Flora O, Selmo E (2006) Isotopic composition of precipitation in
Northern Italy: reverse effect of anomalous climatic events. Journal of Hydrology 329:
471-76.
Longinelli A, Selmo E (2003) Isotopic composition of precipitation in Italy: a first overall
map. Journal of Hydrology 270:75-88.
Maione, U., Machne, G., (1982) Studio sulla formazione e sulla propagazione delle piene del
fiume Tagliamento [Tecnical report on the Tagliamento flood events]. Relazione tecnica,
23 pp.
Hydrogeology and Geochemistry of the Friuli Venezia Giulia Plain Alluvial Aquifers… 257

Spaliviero M (2003) Historic fluvial development of the Alpine-foreland Tagliamento River,


Italy, and consequences for floodplain management. Geomorphology 52, 317–333.
Mosetti F (1983) Sintesi sull’idrologia del Friuli Venezia Giulia [Summary of the Friuli
Venezia Giulia hydrology]. Quaderni Ente Tutela Pesca 6, 1-296
Parkhurst DL, Thorstenson DC, Plummer L N (1980) PHREEQE a computer program for
geochemical calculations. USGS Water Resources Investigations Report, 80- 96
Pilla G (1998) Hydrogeochemical and Geochemical Isotopic characterization of groundwater
resources in the underground of Pavia (South Western Lombardy, Italy). Atti Ticinesi
Scienze della Terra 40:185-201
Pilla G, Sacchi E, Zuppi G, Braga G, Ciancetti G (2006) Hydrochemistry and isotope
geochemistry as tools for groundwater hydrodynamic investigation in multilayer aquifers:
a case study from Lomellina, Po plain, South-Western Lombardy, Italy. Hydrogeology
Journal 14:795-808.
Pilli A (2005) Conceptual flow model of the plain–pre-Alps system in the area between
Vicenza and Trento (Northeastern Italy). PhD, Cafoscari University, Venezia.
Regione Autonoma Friuli-Venezia Giulia, Direzione Centrale Ambiente e Lavori Pubblici
(2007). Le acque calde della Pianura Friulana. Realizzazione della Carta Geologico-
Tecnica della Risorsa Geotermica Regionale e definizione delle Linee Guida per il suo
utilizzo [The geothermal waters of the Friuli Venezia Giulia region]. Regione Friuli-
Venetia Giulia, 27 pp.
Slejko D., Carulli G.B., Nicolich R., Rebez A., Zanferrari A., Cavallin A., Doglioni C.,
Carraro F., Castaldini D., Iliceto V., Semenza E., Zanolla C. (1989) - Seismotectonics of
the eastern Southern-Alps: a review. Bolletino Geofisico Teorico Applicato, 31, 109-136.
Stefanini S, Cucchi F (1976) Gli acquiferi nel sottosuolo della provincia di Gorizia [The
aquifers in the Gorizia province]. Quaderni Isontini Ricerca Acque 28: 347-66.
Stefanini S, Cucchi F (1977) Gli acquiferi nel sottosuolo della provincia di Udine [The
aquifers in the Udine province]. Quaderni Isontini Ricerca Acque 34: 131-47.
Vreča P, Bronci IK, Horvatincic N, Baresic J (2006) Isotopic characteristics of precipitation
in Slovenia and Croatia: comparison of continental and maritime stations. Journal
Hydrology 330:457-69.
Zuppi GM, Sacchi E (2004). Hydrogeology as a climate recorder: Sahara-Sahel (North
Africa) and the Po Plain (Northern Italy). Global Planetary Change 40:79-91.
In: Groundwater: Modelling, Management... ISBN 978-1-60456-832-5
Editors: L.F. Konig and J.L. Weiss, pp. 259-289
c 2008 Nova Science Publishers, Inc.

Chapter 10

A NALYTICAL AND N UMERICAL S OLUTIONS FOR


C ONVECTION -D IFFUSION -D ISPERSION -R EACTION
E QUATIONS WITH G ENERAL I NITIAL -C ONDITIONS :
T HEORY AND A PPLICATIONS
Jürgen Geiser∗
Department of Mathematics,
Humboldt Universität zu Berlin,
Unter den Linden 6, D-10099 Berlin, Germany

Abstract
Our motivation to write this paper came from a real-life model simulating waste-
disposal embedded in an overlying rock. The main problem for our model is the
large time-periods necessary to obtain a realistic scenario to foresee the contamina-
tion process. Because of this multi-scale problem, which occurred due to the coupled
reaction terms of our underlying system of convection-diffusion-dispersion-reaction-
equations, standard solver methods are mostly ineffective to achieve realistic time-
periods. Due to this fact, we developed analytical and numerical methods that al-
lowed a computation over a large simulation period of more than ten thousand years.
We constructed discretization methods of a higher-order with embedded analytical
solutions, which allow large-time-steps without loss of accuracy. Based on spatial-
and time-decomposition methods, we decouple complex equations into simpler equa-
tions and use adequate methods to solve each equation separately. For the explicit
parts that are the convection-reaction-equations we use finite-volume methods based
on flux-methods with embedded analytical solutions. Whereas for the implicit parts
that are the diffusion-dispersion-equations we use finite-volume methods with cen-
tral discretizations. We analyze the splitting-error and the discretization error for our
methods. The main contribution of the paper is the design of analytical solutions
with general initial conditions that can be used to discretize the explicit parts with
mass-conserved higher-order methods. The verification of the new methods is done
for benchmark and real-life applications with our programme tool R3 T . The test-
examples and benchmark problems are taken into account for the general initial con-
ditions and verify our discretization- and solver-methods with respect to the physical

E-mail address: geiser@mathematik.hu-berlin.de
260 Jürgen Geiser

behavior. Based on the verification of realistic model-problems we discuss a waste-


disposal in three dimensions (3D) with large decay-chains reacted and transported in
a porous media with underlying flowing groundwater. For the prediction of possible
waste-disposal a computation with different located waste-locations is discussed.

Keywords. Convection-diffusion-dispersion-reaction equation, operator-splitting, finite-


volume method, embedded analytical solutions, numerical simulations.
AMS subject classifications. 35K15 35K57 47F05 65M60 65N30.

1. Introduction
Our studies are motivated by a desire to model the transport of radioactive and chemical
contaminants through an overlying rock with waste-scenarios of large time-periods. We
are interested, in particular, in large-scale simulations for long time contaminations in the
underlying porous media, see [10]. The mathematical model provides us with a coupled sys-
tem of convection-diffusion-dispersion-reaction equations and arbitrary initial conditions.
To solve such equation systems on different scales, we have to deal with adapted discretiza-
tion and solver methods, see [25] and [18]. One main idea is to decouple the full equations
and solve each simpler equation on the adapted time and domain scales, see [20]. For such
a splitting we apply and develop improved algorithms regarding our target to achieve higher
efficiency and more exact accuracy. The reaction equations, as well as the convection equa-
tions, belong to the small time scale and we use the explicit temporal discretization based
on analytical solutions, see [32] and [9]. The discretization methods are derived for the
convection-reaction equation with respect to finite volume and characteristic methods, see
[4] and [5]. The main contribution is a modified discretization method with embedded an-
alytical one-dimensional solutions in the multi-dimensional finite volume methods. The
analytical derivation is done by Laplace-transformation, see [8], and the solutions can be
generalized by polynomial initial conditions. The embedding into the mass-conserved dis-
cretization method, which is given by the finite volume method, can be achieved with the
Godunov method. By decoupling the multi-dimensional problem in one dimensional prob-
lems, we can apply the accurate analytical solutions. Our work derivates form the work of
Morton [29] and Celia [5], while specializing in the system of convection-reaction meth-
ods for arbitrary initial values. By transferring the problem to a mass transport we further
conserve the underlying mass and have a conservative method [28].
For the larger scales, appertaining equations are the diffusion and dispersion equa-
tions, we use the implicit temporal discretization methods. Higher-order methods for the
convection-reaction part are used by reconstruction with linear one dimensional exact test-
functions and generalized by dimensional splitting methods to multi-dimensional applica-
tions. We obtain for the diffusion-dispersion equation implicit temporal discretization and
the higher-order finite volume methods. The underlying linear system of equations is solved
iteratively with a multi-grid solver. Our main advantage in this context is the time-splitting
method of the different equation types to obtain higher-order discretization methods.
Both results are coupled by a time-splitting method, e.g. Strang-splitting method to
obtain at least second-order in time and second-order in space.
Analytical and Numerical Solutions... 261

Using meliorated higher-order operator-splitting methods, we can improve our methods


for the solution of the full equations.
The methods are verified by benchmark problems and the numerical results are com-
pared with the analytical solutions. General initial conditions are verified by polynomial
initial conditions in an axisymmetric two-dimensional application. A real-life problem is
presented as a simulation of a waste-disposal with realistic parameters and underlying lay-
ers in the porous media. The calculations are presented within the figures and convergence
results. Some applications containing these methods are computed with the underlying
programme tool R3 T , and the main concepts are presented.
The paper is outlined as follows. We introduce our mathematical model of a contami-
nant transport in flowing groundwater in Section 2.. The decomposition methods are pre-
sented in Section 3.. In Section 4. we introduce the finite volume methods to be used as basic
discretization methods to develop higher-order methods based on the analytical solutions.
The analytical solutions with general initial conditions, such as embedded mass-conserved
equations for the finite volume methods, are explained in Section 5.. Our numerical results
with benchmark problems and realistic waste-disposals are described in Section 6.. Finally
we discuss our future works in Section 7. with respect to our research area.

2. Mathematical Model
The motivation for the study presented below comes from a computational simulation of
radionuclide contaminants transported in flowing groundwater [10], [13].
The mathematical equations are given by:
X
∂t Rα (cα ) + ∇ · (vcα − D∇cα ) + λαβ Rα (cα ) = λγα Rγ (cγ ) , (1)
γ∈γ(α)
Rα (cα ) = (φ + (1 − φ)ρK(cα )) cα , (2)
with α = 1, . . . , m . (3)

The unknowns cα = cα (x, t) are considered in Ω × (0, T ) ⊂ IRd × IR, the space-dimension
is given by d. φ is the porosity, ρ is the density. The retardation Rα (cα ) is given as a
linear or nonlinear function. For the linear retardation we have the Henry-Isotherm with
K(cα ) = Kdα , where Kdα is the element-specific Kd -parameter. We simplify the model
and assume that for each element one isotope confers our complex model [18]. For the
nonlinear retardation-factor we have the Freundlich-Isotherm with K(cα ) = Knl (cα )p−1 ,
where Knl is the specific sorption-constant and p is the exponent of the isotherm. Another
κb
nonlinear retardation-factor is the Langmuir-Isotherm with K(cα ) = 1+b cα , where b is the
specific sorption-constant and κ is the specific sorption-capacity. The other parameters are
λαβ , the decaying rates from α to β, where γ(α) are the predecessor elements of element
α. D is the Scheidegger diffusion-dispersion tensor and v is the velocity.
The main aim of this paper is to present new methods, which are based on exact solu-
tions for simpler one-dimensional equations. For this purpose we derive analytical solutions
for the simpler one-dimensional convection-reaction equation and also for the nonlinear re-
action equations. The analytical solutions are used for the explicit time-discretization and
spatial-discretization with finite volume methods for d-dimensions.
262 Jürgen Geiser

A higher-order method for the linear convection-reaction equation is derived with the
idea to embed the analytical solution of the mass to our finite volume discretization. The
mass-transport coupled the convection-reaction equation together. For the nonlinear re-
tardation our methods are based on the higher-order operator-splitting methods. We could
couple analytical solutions for the nonlinear reaction equation with the convection-diffusion
equation together. For these coupling methods we also get higher-order methods, which al-
low larger, coarser grids with larger time-steps. We could therefore fulfill the forced long
time-period calculations, e.g. scenarios about ten thousand years are calculatable in one
day.
The higher-order finite volume method is based on TVD methods and is constructed
according to the discrete minimum and maximum principle that is used by the new methods
to reach second-order for all components.
For the derivation of our new methods based on the embedding of one-dimensional
analytical solutions for convection-reaction equations we use the d dimensional convection-
reaction equations with equilibrium sorption.
So the equation for this linear case is given by:

Ri ∂t ci + ∇ · (vci ) = −λi Ri ci + λi−1 Ri−1 ci−1 , i = 1, . . . , m , (4)

ci = (c1,i , . . . , cd,i )T ∈ IRd ,


where Ri ≥ 0 is a constant retardation factor. The trivial inflow and outflow boundary con-
ditions are given by c = 0. The polynomial initial conditions ci (x, 0) = ci,0 (x) are given
as rectangular-, trapezoidal- and also piecewise-linear-impulses. Such polynomial initial
conditions make it possible to derive the analytical solutions for a one-dimensional domain.
Based on the one-dimensional convection-reaction equation with equilibrium sorption and
general initial impulses, we derive new discretization methods, see [18]. A generalization to
multi-dimensional domains can be achieved by decoupling into one-dimensional domains
with respect to spatial decomposition methods.
The following section describes the time and spatial decomposition methods as a basic
method for our new discretization methods.

3. Time and Spatial Decomposition Methods


The time- and spatial decomposition methods are used to solve complex models in geo-
physical and environmental physics, they were developed and applied in the 1990s, see [6],
[33], [34].

3.1. Spatial Decomposition Method Based on Overlapping Schwarz Wave-


Form Relaxation
We present the overlapping Schwarz wave-form relaxation method as spatial decomposi-
tion method for the solution of the convection-diffusion-reaction equation with constant
coefficients, see [6]. We utilize the decomposition for arbitrary multi-dimensional domains
and elaborate the impact of the coupling on the convergence of the overlapping Schwarz
wave-form relaxation.
Analytical and Numerical Solutions... 263

The following model problem is given as an example:

ut + Au = f , in Ω × (0, T ) , Ω × (0, T ) := Ω1 × (0, T ) ∪ Ω2 × (0, T ) , (5)


u(x, 0) = u0 , (6)
u = g , on ∂Ω × (0, T ) , (7)

where A denotes, for example, a second-order partial differential operator Au =


−∇D∇u + v∇u + cu for the given coefficients D ∈ IR+ , v ∈ IRn , c ∈ IR+ , and n is
the dimension of the space or a first-order partial differential equation as given in our model
equation (1) in Section 2.. The underlying domains Ω1 and Ω2 are convex and Lipschitzian
and do not influence the following analysis. Each iteration step consists of two half-steps
associated with the two sub-domains and we solve two subproblems:

u1t + Aun1 = f , in Ω1 × (0, T ) ,


u1 (x, 0) = u10 ,
(8)
un1 = un−1
2 , on L2 = ∂Ω1 × (0, T )\∂Ω × (0, T ) ,
un1 = g , on L0 = ∂Ω × (0, T ) ∩ ∂Ω1 × (0, T ) ,

u2t + Aun2 = f , in Ω2 × (0, T ) ,


u2 (x, 0) = u20 ,
(9)
un2 = un1 , on L1 = ∂Ω2 × (0, T )\∂Ω × (0, T ) ,
un2 = g , on L3 = ∂Ω × (0, T ) ∩ ∂Ω2 × (0, T ) .

Remark 1. For the non-overlapping sub-domains the values at the interfaces are predicted
by the use of an explicit scheme and the problem is solved over each sub-domain indepen-
dently. This method is of the non-iterative type but it has a drawback regarding the stability
condition for interface prediction by the explicit method and the solution by the implicit
scheme or any other unconditional stable finite difference scheme [7].

3.2. First Order Time-Splitting Methods


The following splitting methods of first-order are described. We consider the following
ordinary linear differential equations:

∂t c(t) = A c(t) + B c(t) , (10)

whereby the initial-conditions are cn = c(tn ) . The operators A and B are spatial-
discretised equations, e.g. the convection equation and the diffusion equation.
The operator-splitting method is a method which solves the two operators sequentially,
with respect of the initial conditions. We get two sub-equations :

∂c∗ (t)
= Ac∗ (t) , with c∗ (tn ) = cn , (11)
∂t
∂c∗∗ (t)
= Bc∗∗ (t) , with c∗∗ (tn ) = c∗ (tn+1 ) .
∂t
whereby the time-step is τ n = tn+1 − tn . The solution of equations are cn+1 = c∗∗ (tn+1 ).
264 Jürgen Geiser

The error of this splitting method is derived as


1
ρn = (exp(τ n (A + B)) − exp(τ n b) exp(τ n A)) c(tn )
τ
1 n
= τ [A, B] c(tn ) + O(τ 2 ) . (12)
2
whereby [A, B] := AB − BA is the commutator of A and B. We get an error O(τ n ) if the
Operators A and B do not commutate, otherwise the method is exact.

3.3. Higher-Order Time-Splitting Methods


We could improve our method by the so-called Strang-splitting method, which is of second-
order, see [31].
The method is presented as:
∂c∗ (t)
= Ac∗ (t) , with tn ≤ t ≤ tn+1/2 and c∗ (tn ) = cn , (13)
∂t
∂c∗∗ (t)
= Bc∗∗ (t) , with tn ≤ t ≤ tn+1 and c∗∗ (tn ) = c∗ (tn+1/2 ) ,
∂t
∂c∗∗∗ (t)
= Ac∗∗∗ (t) , with tn+1/2 ≤ t ≤ tn+1 and c∗∗∗ (tn+1/2 ) = c∗∗ (tn+1 ) ,
∂t
whereby the results of the method is cn+1 = c∗∗∗ (tn+1 ).
The splitting error of this method is given as, confer [24]:
1 n 2
ρn = (τ ) ([B, [B, A]] − 2[A, [A, B]]) c(tn ) + O((τ n )4) . (14)
24
whereby we get the second-order if the operators do not commute and an exact result if they
do commute.
We could improve the order by using more intermediate steps. A further higher-order
splitting method is the following iterative method.

3.4. The Iterative Time-Splitting Method


The following algorithm is based on the iteration with fixed splitting discretization step-
size τ . On the time interval [tn , tn+1 ] we solve the following subproblems consecutively
for i = 0, 2, . . . 2m.

∂ui (x, t)
= Aui (x, t) + Bui−1 (x, t), with ui (tn ) = un (15)
∂t
u0 (x, tn ) = un , u−1 = 0,
and ui (x, t) = ui−1 (x, t) = u1 , on ∂Ω × (0, T ) ,
∂ui+1 (x, t)
= Aui (x, t) + Bui+1 (x, t), (16)
∂t
with ui+1 (x, tn ) = un ,
and ui (x, t) = ui−1 (x, t) = u1 , on ∂Ω × (0, T ) ,
Analytical and Numerical Solutions... 265

where un is the known split approximation at the time level t = tn (see [?]).

Remark 2. We can generalize the iterative-splitting method to a multi-iterative splitting


method by introducing new splitting operators, e.g. spatial operators. Then we obtain multi-
indices to control the splitting process, each iterative splitting method can be solved inde-
pendently, while connecting with further steps to the multi-splitting methods. In the follow-
ing we introduce the multi-iterative-splitting method for a combined time-space-splitting
method.
In the next section we present the discretization methods for the equations.

4. Discretization
For the space-discretization we use finite volume methods and for the time-discretization
we use explicit or implicit Euler methods. In the next section we introduce the notation
for the space-discretization. Further, for the equation’s terms we describe the discretization
methods.

4.1. Notation
The time-steps for the calculation in the time-intervals are (tn , tn+1 ) ⊂ (0, T ) , for n =
0, 1, . . .. The computational cells are given as Ωi ⊂ Ω with j = 1, . . . , I. The unknown I
is the number of nodes.
For the use of finite volumes we have to construct the dual mesh for the triangula-
tion T [15] of the domain Ω. First, the finite elements for the domain Ω are given by
T e , e = 1, . . . , E. The polygonal computational cells Ωj are related to the vertices xj of
the triangulation.
To get the relation between the neighbour cells and to use the volume of each cell we
introduce the following notation. Let Vj = |Ωj | and the set Λj denotes the neighbour-point
xk to the point xj . The boundary of cells j and k are Γjk .
We define the flux over the boundary Γjk as
Z
vjk = n · v ds . (17)
Γjk

The inflow-flux is given as vjk < 0, the outflow-flux is vjk > 0. The fluxes’ antisymmetrics
are denoted as vjk = −vkj . The total outflow-flux is given by:
X
νj = vjk . (18)
k∈out(j)

The idea of the finite volumes is to construct an algebraic equation system to express
the unknowns cni ≈ c(xi , tn ). The initial values are given with c0i . The expression of
the interpolation schemes could be given naturally in two ways, the first is given with the
primary mesh of the finite elements:
I
X
cn = cni φi (x) (19)
i=1
266 Jürgen Geiser

where φi are the standard globally-finite-element-basis functions [15]. The second possi-
bility is for the finite volumes with:
I
X
ĉn = cni ϕi (x) (20)
i=1

where ϕi are piecewise constant discontinuous functions defined by ϕi (x) = 1 for x ∈ Ωi


and ϕi (x) = 0 otherwise.

4.2. Discretization of the Convection Equation with First-Order


The piecewise constant discretization of the convection equation

∂t R c − v · ∇c = 0 , (21)

with the simple boundary condition c = 0 for the inflow and outflow boundary. We use the
upwind discretization done in [15] and get
X
Vj R cn+1
j = cnj (R Vj − τ n νj ) + τ n R cnk vkj , (22)
k∈in(j)

Because of the explicit discretization for time and to fulfill the discrete minimum-maximum
property [15], we get a restriction for the time-steps as follows:
R Vj
τj = , τ n ≤ min τj (23)
νj j=1,...,I

In the next subsection we improve the discretization with a reconstruction with linear
polynomes. The reconstruction is based on Godunov’s method and is the limiter on the
local minimum-maximum property.

4.3. Discretization of the Convection Equation with Higher-Order


The reconstruction was done in a previous paper [15] and we introduce the scheme in the
next steps.
The linear polynomes are reconstructed over the element-wise gradient and are given
as:

un (xj ) = cnj , (24)


E Z
1 X
∇un |Vj = ∇cn dx , (25)
Vj T e ∩Ω j
e=1
with j = 1, . . . , I .

The piecewise linear function is given by:

unjk = cnj + ψj ∇un |Vj (xjk − xj ) , (26)


with j = 1, . . . , I ,
Analytical and Numerical Solutions... 267

where ψi ∈ (0, 1) is the limiter which has to fulfill the discrete minimum-maximum prop-
erty.
The piecewise linear function is given by:

unjk = cnj + ψj ∇un |Vj (xjk − xj ) , with j = 1, . . . , I ,

where ψj ∈ (0, 1) is the limiter which has to fulfill the discrete minimum-maximum prop-
erty, as described in [15].
We also use the limitation of the flux to get no overshooting, when transporting the mass
and to receive the maximal time-step.
We get the restriction for the concentration as:
τj n
ũnjk = unjk + (c − unjk ) . (27)
τn j
Using all the previous schemes the discretization for the second-order is written in the
form
X X
RVj cn+1
j = RVj cnj − τ n ũnjk vjk + τ n ũnlj vlj , (28)
k∈out(j) l∈in(j)

This discretization method is used for the next coupled discretization with the reaction
equation.

4.4. Discretization of the Convection-Reaction Equation


with One-Dimensional Analytical Solutions
We apply Godonov’s method for the discretization, see [28], and enlarge it with the solution
of convection-reaction equations. We reduce the equation to a one-dimensional problem,
solve the equation exactly and transform the one-dimensional mass to the multi-dimensional
equation.
The discretization of the equation:

∂t cl + ∇ · vl cl = −λl cl + λl−1 cl−1 , (29)


with l = 1, . . . , m .

The velocity vector v is normed by Rl . The initial conditions are given by c01 = c1 (x, 0)
, else c0l = 0 for l = 2, . . . , m and the boundary conditions are trivial cl = 0 for l =
1, . . . , m.
We first calculate the maximal time-step for cell j and concentration i with the use of
the total outflow fluxes:
V j Ri X
τi,j = , νj = vij .
νj
j∈out(i)

We get the restricted time-step with the local time-steps of cells and their components:

τ n ≤ min τi,j .
i=1,...,m
j=1,...,I
268 Jürgen Geiser

The velocity of the discrete equation is given by:


1
vi,j = .
τi,j
We calculate the analytical solution of the mass with equation (29), see also Subsection
5.1., and obtain:

mni,jk,out = mi,out (a, b, τ n , v1,j , . . . , vi,j , R1 , . . . , Ri , λ1 , . . . , λi ) ,


mni,j,rest = mni,j f (τ n , v1,j , . . . , vi,j , R1 , . . . , Ri , λ1 , . . . , λi ) ,

whereby a = Vj Ri (cni,jk − cni,jk′ ) , b = Vj Ri cni,jk′ and mni,j = Vj Ri uni are the parameters.
The linear impulse in the finite-volume cell is cni,jk′ for the concentration on the inflow- and
cni,jk for the concentration on the outflow-boundary of the cell j.
The discretization with the embedded analytical mass is calculated by:
X vjk X vlj
mn+1
i,j − m n
i,rest = − m i,jk,out + mi,lj,out ,
νj νl
k∈out(j) l∈in(j)

vjk
whereby νj is the re-transformation for the total mass mi,jk,out in the partial mass mi,jk
. The mass in the next time-step is mn+1
i,j = Vj ci
n+1
and in the old time-step it is the rest
mass for the concentration i. The proof is done in [18]. In the next section we derive an
analytical solution for the benchmark problem.

4.5. Discretization of the Reaction Equation


The reaction equation is an ordinary-differential equation and is given as follows:

∂t Ri ci = −λi Ri ci + λi−1 Ri−1 ci−1 , (30)

whereby i = 1, . . . , m and λ0 = 0 is. The decay factors are λi ≥ 0.0 and the retardation
factors are Ri > 0.0. The initial conditions are c1 (x, t0 ) = c01 and ci (x, t0 ) = 0 with
i = 2, . . . , m.
We could derive the solutions for these equations, see [3], with:
i
R1 X
ci = c01 Λi Λj,i exp(−λj t) , (31)
Ri
j=1

whereby i = 1, . . . , m. The solutions are defined for all λj 6= λk for j 6= k and j, k ∈


1, . . . , M .
The factors Λi and Λj,i are
i−1 i
Y Y 1
Λi = λj , Λj,i = . (32)
λk − λj
j=1 j=1
j6=k

For pairwise equal-reaction factors we have derived the solution in our work [18].
In the next subsection we introduce the discretization of the diffusion-dispersion equa-
tion.
Analytical and Numerical Solutions... 269

4.6. Discretization of the Diffusion-Dispersion Equation


We discretize the diffusion-dispersion equation with implicit time-discretization and finite
volume method for the following equation:
∂t R c − ∇ · (D∇c) = 0 , (33)
whereby c = c(x, t) with x ∈ Ω and t ≥ 0 . The diffusions-dispersions tensor is D =
D(x, v) given by the Scheidegger approach, see [30]. The velocity is given by v. The
retardation factor is R > 0.0.
We have the boundary values with n · D ∇c(x, t) = 0, whereby x ∈ Γ is the boundary
Γ = ∂Ω, see [14].
We integrate the equation (33) over space and time and get:
Z Z tn+1 Z Z tn+1
∂t R(c) dt dx = ∇ · (D∇c) dt dx . (34)
Ωj tn Ωj tn

The integration over time is done with the backward Euler method and the lumping for the
diffusion-dispersion term, see [18]:
Z Z
(R(cn+1 ) − R(cn )) dx = τ n ∇ · (D∇cn+1 ) dx , (35)
Ωj Ωj

The equation (35) is discretised over the space using the Green’s formula.
Z Z
n+1 n n
(R(c ) − R(c )) dx = τ D n · ∇cn+1 dγ , (36)
Ωj Γj

whereby Γj is the boundary of the finite volume cell Ωj . We use the approximation in
space, see [18].
The integration for the equation (36) is done for finite boundary and by the use of the
middle-point rule:

∇ce,n+1
X X
Vj R(cn+1
j ) − Vj R(cnj ) = τ n |Γejk |nejk · Djk
e
jk , (37)
e∈Λj k∈Λej

whereby |Γejk | is the length of the boundary-element Γejk . The gradients are calculated with
the piecewise finite-element function φl , see equation (37) and we get

∇ce,n+1
X
jk = cn+1
l ∇φl (xejk ) . (38)
l∈Λe

We get with the difference notation for the neighbor point j and l, see [16], in the
following discretization form:
Vj R(cn+1
j ) − Vj R(cnj ) = (39)
X X X 
= τn e
|Γejk |nejk · Djk ∇φl (xejk ) (cn+1
j − cn+1
l ),
e∈Λj l∈Λe \{j} k∈Λej

whereby j = 1, . . . , m.
In the next section we introduce the analytical solutions for the discretization methods
of the convection-reaction equation.
270 Jürgen Geiser

5. Analytical Solutions
For the next section we deal with the following system of one-dimensional convection-
reaction equations without diffusion. The equation is given as:

∂t ci + vi ∂x ci = −λi ci + λi−1 ci−1 , (40)

for i = 1, . . . , M . The unknown M is the number of components. The unknown functions


ci = ci (x, t) denote the contaminant concentrations. They are transported with piecewise
constant (and in general different) velocities vi . They decay with constant reaction rates λi .
The space-time domain is given by (0, ∞) × (0, T ).
We assume a simple (irreversible) form of decay chain, e.g. λ0 = 0 and for each
contaminant only single source term λi−1 ci−1 is given. For simplicity, we assume that
vi > 0 for i = 1, . . . , M .
We describe the analytical solutions with piecewise linear initial conditions. But all other
piecewise polynome functions could be derived, see [18].
For boundary conditions we take zero concentrations at inflow boundary x = 0 and the
initial conditions are defined for x ∈ (0, 1) with:

ax + b , x ∈ (0, 1)
c1 (x, 0) = ,
0 otherwise (41)
ci (x, 0) = 0 , i = 2, . . . , M ,

where a, b ∈ IR+ are constants.


We use Laplace Transformation for the transformation of the partial differential equa-
tion into the ordinary differential equation, described in [21]. We solve the ordinary dif-
ferential equations, described in [9], and retransform the solution in the original space of
the partial differential equations. We could then use the solution for the one-dimensional
convection-reaction equation, see [18].
The solutions are given as:

 0 0 ≤ x < v1 t
c1 (x, t) = exp(−λ1 t) a(x − v1 t) + b v1 t ≤ x < v1 t + 1 , (42)
0 v1 t + 1 ≤ x

 
i i
X X
ci (x, t) = Λi  exp(−λj t)Λj,i Λjk,i Ajk  , (43)

j=1 k=1
k6=j


 0 0 ≤ x < vj t




 a(x − vjat)



Ajk = +(b − λjk )(1 − exp(−λjk (x − vj t))) vj t ≤ x < vj t + 1 . (44)



a

 (b − λjk + a) exp(−λjk (x − vj t − 1))



 −(b − a ) exp(−λ (x − v t))

vj t + 1 ≤ x
λjk jk j
Analytical and Numerical Solutions... 271

whereby the general solutions have the following definition-array:


vi 6= vj , λi 6= λj , λij 6= λik and vi 6= vj ∧ λi 6= λj ,
∀ i, j, k = 1, . . . , M , if i 6= j ∧ i 6= k ∧ j 6= k.
The further abbreviation for λjk and Λi are:

i−1
λj − λk Y
λkj = λjk := , Λi := λj , (45)
vj − vk
j=1

and the factors Λj,i and Λjk,i are:

   
i i
Y 1  Y 1 
Λj,i =   , Λjk,i =  . (46)
k=1
λk − λj 
l=1
λ k − λ j

k6=j l6=j
l6=k

The solutions (42) and (43) are used in the discretization methods for the embedded ana-
lytical mass. To complete the study for the analytical solutions, we introduce the special
solutions for the equal reaction-parameter λl = λa(l) .

5.1. Mass Reconstruction

To calculate the analytical method correctly in a numerical method based on mass exchange,
we derived the analytical mass for the given equations.
The modified method uses the analytical mass for the computation. Now it has con-
structed the analytical mass.
The total mass is given by

mi,sum = mi,rest (t) + mi,out (t) (47)

where mi,rest is the mass, that rests in the interval (0, 1) and mi,out is the mass, that flows
of the interval (1, ∞).
The cells are computed with the following equations over the one-dimensional
Pi cell
(0, 1). For the piecewise independent solution that is given by ci = c
j=1 ij we could
calculate mi,rest by:

Z 1
mi,rest = ci (x, t)dx (48)
0
Z 1 i
X
= cij (x, t)dx (49)
minik=1 {vk t} j=1
272 Jürgen Geiser

The mass is the integrated:


 
i−1 i Yi i i
Y X 1 X Y λjl 
mi,rest = λj  ( ) ( ) exp(−λj t)
 
 λk − λj k=1 l=1 λjl − λjk 
j=1 k=1j=1
k6=j k6=j l6=k
l6=j

1  
a
Z
a(x − vj t) + (b − )(1 − exp(−λjk (x − vj t))) dx
vj t λjk
for i = 2, . . . , M
(50)

This form could be written as a compressed equation, using the following four expressions.
a.) The polynomes with the constant factors are reduced to:
i i
X Y λjl
( )1 = 1 (51)
k=1 l=1
λjl − λjk
k6=j l6=k
l6=j

b.) The polynomes with the linear factor are reduced to:
i i i
X Y λjl 1 X 1
( ) = (52)
k=1 l=1
λjl − λjk λjk k=1
λjk
k6=j l6=k k6=j
l6=j

c.) The polynomes with the quadratical factors are reduced to:
i i i i
X Y λjl 1 X 1 X 1
( ) = ( ) (53)
k=1 l=1
λjl − λjk (λjk )2 k=1
λjk l≥k λjl
k6=j l6=k k6=j l6=j
l6=j

d.) The symmetry of the terms with the exp expressions are from the form of:

exp(−λj t) exp(λjk (1 − vj t)) = exp(−λk t) exp(−λkj (1 − vk t)) (54)

By using with the four reductions we could write the equation in an applicable form:
i−1 i Y i
Y X 1
mi,rest (t) = λj ( ) (55)
λk − λj
j=1 j=1 k=1
k6=j

i
 (1 − vj t)2 X 1
exp(−λj t) a + b(1 − vj t − )
2 k=1
λjk
k6=j

i Xi i 
X 1 1 X 1
−a(1 − vj t)( )+a ( ) 

k=1
λjk k=1
λjk l≥k λjl
k6=j k6=j l6=j
Analytical and Numerical Solutions... 273

The outflowing mass mi,out in the finite cell is then given by the difference of the total
mass mi,sum and the resting mass mi,rest .
So it is important to compute the total mass, so that we could prove the mass is given
as the solution with the ordinary equation and the volume of the initial condition. We could
order the velocities with the notation of the permutation vk(1) < . . . < vk(i) , and without a
restriction we use the following order v1 < v2 < . . . < vi . So we could compute the mass:
Z 1+maxij=1 (vj t)
mi,sum (t) = ci (x, t)dx (56)
minij=1 (vj t)
i Z
X 1+vj t i−1 Z
X 1+vi t
= cj (x, t)dx + cj (x, t)dx
j=1 vj t j=1 1+vj t

The integration fulfilled the equation:


i−1 i Y i i i
R1 Y X 1 X Y λjl
mi,sum (t) = λj ( ( ) ( ) exp(−λj t)
Ri λk − λj k=1 l=1 λjl − λjk
j=1 j=1
k=1
k6=j k6=j l6=k
l6=j

1 a 1
(a + (b − )(1 + (exp(−λjk ) − 1)))
2 λjk λjk
i−1 i−1 Yi i i
R1 Y X 1 X Y λjl
+ c01 λj ( ( ) ( ) exp(−λj t)
Ri λk − λj k=1 l=1 λjl − λjk
j=1 j=1
k=1
k6=j k6=j l6=k
l6=j

a 1
((b − + a)(− (exp(−λjk (vi − vj )t) − 1))
λjk λjk
a 1
−(b − )(− (exp(−λjk ((vi − vj )t + 1)) − exp(−λjk )))
λjk λjk

With the reduction of the four concentrations given in (51)-(54) we could reduce to the
equation:
 
i−1 i Y i
R1 Y X 1 1
mi,sum (t) = λj  ( ) exp(−λj t)(a + b)
 
Ri λ k − λ j 2
j=1 k=1j=1
k6=j

But this equation is an ordinary-differential equation and we could reduce the applica-
tion to much simpler notations.
This equation of the analytical mass is then embedded in the discretization which is
introduced as follows.

5.2. Mass Reconstruction for Equal-Decay Factors


The case for two components with equal-decay factors are described. We derived the ana-
lytical solution for two equal-decay factors, see [18].
The solution for the equations are:
274 Jürgen Geiser


 (b + a(x − v1 t)) e−λ1 t x ∈ (v1 t , 1 + v1 t)
u1 (x, t) = (57)
 0 otherwise

α1 α2
u2 (x, t) = λ1 exp(−λ1 t)( + ) (58)
v2 − v1 v1 − v2
 0 0 ≤ x ≤ vl t

(x−vl t)2
αl = a + b(x − vl t) vl t ≤ x ≤ vl t + 1 (59)
 1 2
a2 + b otherwise

We derive the mass for the two parts of the concentration, m21 , m22 and m2,ges =
m21 + m22 .
We integrate the mass m21 :
Z 1 Z 1
m21 (t) = u21 (x, t) dx + u21 (x, t) dx (60)
v1 t v2 t
1
1 a(x − v1 t)2
Z
= λ1 exp(−λ1 t) ( + b(x − v1 t)) dx (61)
v1 t v2 − v1 2
1
1 a(x − v2 t)2
Z
+ ( + b(x − v2 t)) dx (62)
v 2 t v1 − v2 2
 1 a(1 − v1 t)3 b(1 − v1 t)2 
= λ1 exp(−λ1 t) + (63)
v2 − v1 6 2
1 a(1 − v2 t)3 b(1 − v2 t)2  
+ + (64)
v1 − v2 6 2
(65)

We then calculate the total mass m2,ges as given for equal-decay factors in the work
[18] :
1 R1
m2,ges (t) = ( a + b) λ1 t exp(−λ1 t) . (66)
2 R2

5.3. Analytical Solutions of Equilateral Triangle Initial Conditions (Bench-


mark Problem)
To define a benchmark problem, we derive an exact solution for a triangular initial condi-
tion. For this initialization the theory of the convergence rates predicate an order of two,
because of the continuous function.
To start with the boundary conditions we take zero concentrations at inflow boundary
x = 0 and the initial conditions are defined for x ∈ (0, 1):

x ∈ (0, 2ǫ )

 x,
2
c1 (x, 0) = c0 ǫ ǫ − x , x ∈ ( 2ǫ , ǫ)
(67)
0 otherwise

ci (x, 0) = 0 , i = 2, . . . , I .
Analytical and Numerical Solutions... 275

where c0 > 0 is the initial concentration and ǫ > 0 is the length of the triangular impulse.
The initial conditions are transferred as follows:

2 (1 − exp(− 2ǫ s))2
û1 (s, 0) = c0 (68)
ǫ s2

The general situation with piecewise polynomial initial conditions could derive and are
shown in [18]. We will only derive the triangular impulse for the benchmark problem.
The transferred initial conditions given by the equation (68) are inserted in the Laplace-
transformed equation, see [18]. We did the re transformation and the solution is given by:

 0 0 ≤ x < v1 t
x − v1 t + b v1 t ≤ x < v1 t + 2ǫ

2 
u1 (x, t) = c0 exp(−λ1 t) (69)
ǫ v t + ǫ − x v1 t ≤ x < v1 t + ǫ
 1


0 v1 t + ǫ < x

i−1 i i i
Y X Y 1 X
ui (x, t) = λj  exp(−λj t)( ) Ajk Λjlk (70)
λk − λj k=1
j=1 j=1 k=1
k6=j k6=j

with i = 2, . . . , n
i
Y λjl
Λjk =
l=1
λjl − λjk
l6=k
l6=j

 0 0 ≤ x < vj t
x − vj t + λ1jk (−1 + exp(−λjk (x − vj t))) ǫ




 v j t ≤ x < vj t + 2
 1
 vj t + ǫ − x + λjk (exp(−λjk (x − vj t))


Ajk = −2 exp(−λjk (x − (vj t + 2ǫ ))) + 1) vj t + ǫ
2 < x < vj t + ǫ
1
λjk (exp(−λjk (x − vj t))





ǫ
 −2 exp(−λjk (x − (vj t + 2 )))



ǫ
+ exp(−λjk (x − (vj t + ǫ)))) vj t + < x < vj t + ǫ

2
with λjk = λkj

5.4. Generalization of the Analytical Methods to Characteristics

In this subsection we generalize the one-dimensional analytical solutions for the


convection-reaction equations for piecewise polynomial characteristics in two dimensions.
The contribution here is the transformation of the two-dimensional characteristics to a one-
dimensional arc-length which has piecewise constant velocities.
For the characteristics we assume a constant velocity between the start- and end-points
of the line. The piecewise constant velocities are given as:
 
f (x(t), y(t), t)
v= , (71)
g(x(t), y(t), t)
276 Jürgen Geiser

where the absolute value of the constant velocities is:


p
v = |v| = f (x(t), y(t), t)2 + g(x(t), y(t), t)2 = const. , (72)
+
∀(x, y) ∈ Ω , ∀t ∈ IR ,
with v ∈ IR+ .

The parameterization of the line can be described as:


 
x(t)
x(t) = , (73)
y(t)

The length or arc-length of the line is given as:


Z t2 p
xarc = (x′ (t))2 + (y ′ (t))2 dt , (74)
t1

where x′ (t) = ∂x(t) ′ ∂y(t)


∂t and y (t) = ∂t is given.
Based on this transformation, we achieve a one-dimensional domain and can apply our
one-dimensional solutions, see (69), (70).
For the triangular-impulse as initial condition we can compute the following function,
see equation (69) and (70):

ui,T ri = ui,T ri (xarc , t, ǫ, c0 , v1 , . . . , vi , λ1 , . . . , λi ) , (75)

where vi = Rvi are the velocities, xarc is the arc-length of the line, ǫ is the length of the
triangular impulse, c0 is the constant value of the initial-conditions, Ri are the retardation
factors and λi are the decay-rates with i = 1, . . . , M .
Based on these assumptions we can generalize the analytical solutions for characteris-
tics on two-dimensional domains.
The transport of the impulses as initial conditions on a characteristic is shown in the
Figure 1.

v1
1
0
0
1
1
0
x arc,1 0
1
0
1 1
0 1
0
1
0 1
0
1
0 v2 1
0
11
00
11
00
11
00
00
11
x arc,2
001
110
0
1
0
1

Figure 1. Two-dimensional model domain: characteristics with constant velocities and


computable arc-lengths.
Analytical and Numerical Solutions... 277

Remark 3. We can generalize the analytical solution to two-dimensional model-problems


with computable arc-lengths. This benefits the discretization methods to achieve higher-
order methods for piecewise polynomial characteristics.

6. Numerical Experiments
In all the following methods we will compare the standard and the modified ones.
The standard method is based on an operator-splitting method which decouples the
convection and reaction term, see the discretization methods in [14].
The modified method is based on the previous section carried out with a coupled method
which is combined with the analytical solution of the mass. This method has a higher-order
discretization because of the embedded analytical solutions for the convection-reaction
equation.
We will next present the experiments done with the standard and modified methods, for
the convection-reaction equation and also for the convection-diffusion-dispersion-reaction
equation. We start with the one- and two-dimensional problems which we compared
with analytical solutions. At the end we present recent experiments done with parallel
computing in two and three dimensions with the full equation.

6.1. First Experiment: Transported Triangle (One-Dimensional Problem)


For the first experiment we use a one-dimensional benchmark problem with delta initial
conditions. The analytical solution is given by equation (69) and (70) which compare the
analytical solution with the numerical solutions.
We calculate the solutions on a two dimensional domain, for which the velocity field
is constant in the x-direction with the constant value of v = (1.0, 0.0)T . We use only the
convection-reaction equation with four components, given in the form

Ri ∂t ui + v∂x ui = −Ri λi ui + Ri−1 λi−1 ui−1 , (76)


i = 1, . . . , 4 . (77)

with the inflow/outflow boundary conditions with n·v ci = 0.0, with no inflow and outflow.
The initial condition is given for the first component with:

−x + 1 0 ≤ x ≤ 1
u1 (x, 0) = (78)
0 otherwise ,
ui (x, 0) = 0.0 , i = 2, . . . , 4 . (79)

For the one-dimensional problem we could compare the numerical solutions with the an-
P sections. We use the L1m
alytical solutions derived in the previous -norm to compare the
solutions, which is given with ELl 1 := i=1,...,I Vi |cm
i − C(x ,
i iy , t )|.
The model domain is given with a rectangle of 8 × 1 units. The initial coarse grid
is given with eight quadratic-unit elements, the uniform refinements are up to the level 7
(131072 Elements).
278 Jürgen Geiser

We choose the parameters to get results at the end of the same maximum value, so that
we would not see the influence of numerical effects with different scalars.
For the first test we use the following parameters:
We use the decay-rates of λ1 = 0.4, λ2 = 0.3, λ3 = 0.2, λ4 = 0.0 and the retardation
factors R1 = 1.0, R2 = 2.0, R3 = 4.0, R4 = 8.0.
The model time is done from t = 0, . . . , 6.0. We compared the results at the end-
time t = 6.0. To do this we compared the L1 -norm and the numerical convergence-rate
ρ = (log(ELl 1 ) − log(ELl−1
1
))/ log(0.5) for the computed levels l = 4, . . . , 7.
So the first results are presented with the standard method and the values are given in
the next table. The L1 -error and the convergence-rate is given as follows:

Table 1. L1 -error for the ascending-retardation factors with the standard method.

l EL1 1 ρ1L1 EL2 1 ρ2L1 EL3 1 ρ3L1 EL4 1 ρ4L1


4 0.0 1.71 10−3 1.04 10−3 2.407 10−4
5 0.0 ∞ 8.61 10−4 0.989 5.28 10−4 0.978 1.22 10−4 0.98
6 0.0 ∞ 4.29 10−4 1.005 2.65 10−4 0.995 6.13 10−5 0.993
7 0.0 ∞ 2.14 10−4 1.003 1.31 10−4 1.016 3.07 10−5 0.997

To compare these results with the modified version we reproduce the same calculations
and get the following results: See in these results we improved the convergence rates for the

Table 2. The L1 -error and the convergence-rate for the ascending-retardation factor
done with the modified method

l EL1 1 ρ1L1 EL2 1 ρ2L1 EL3 1 ρ3L1 EL4 1 ρ4L1


4 0.0 3.06 10−4 3.91 10−5 7.79 10−6
5 0.0 ∞ 8.03 10−5 1.95 9.87 10−6 1.986 2.15 10−6 1.89
6 0.0 ∞ 2.007 10−5 2.0 2.60 10−6 1.93 5.81 10−7 1.89
7 0.0 ∞ 4.36 10−6 2.21 6.66 10−7 1.96 1.51 10−7 1.94

modified results. They tend to the second-order so that we achieve a second-order method
for such results.
For the next test example we use a descending decay chain to get the mixed combina-
tion of the different retardation parameters. The parameters given for the decay-chain are
with the values:
λ1 = 0.3, λ2 = 0.4, λ3 = 0.5, λ4 = 0.0.
The descending-retardation factors are given with the values:
R1 = 8.0, R2 = 4.0, R3 = 2.0, R4 = 1.0

The results of the calculations are given as:


The modified example has the results:
Analytical and Numerical Solutions... 279

Figure 2. Concentration for the 4 components with ascending-retardation factors at time


t=6.

The results also show the improved calculations of the modified method.
To remember the initialization we use the range of the impulse for the first component
until the finite subvolume, because of the discrete values for the initializing. For this kind
we have improved values.
For the next experiment we will compare an original two dimensional problem.

6.2. Second Experiment: Rotating Pyramid


For further applications of two-dimensional problems we focus on a new benchmark prob-
lem. The problem is described in the literature as a rotating Gaussian impulse, see [14].
The application of our test example modified this benchmark problem in a form to use
for the analytical one-dimensional solution.
The constant velocities are given on the circular lines and the continuous form of the
impulse is given with a triangular impulse on the circular arc. The further directions are
continuous on the radius r with a linear function of r beginning in the basic radius ra and
rb as presented in figure 4. So the two-dimensional example is continuous in α- and in
r-direction.
The transformation brings the two-dimensional problem to a one-dimensional one, for
which we derived the analytical solution.
280 Jürgen Geiser

Table 3. L1 -error and convergence rate for the descending-retardation factors for the
standard method.

l EL1 1 ρ1L1 EL2 1 ρ2L1 EL3 1 ρ3L1 EL4 1 ρ4L1


4 7.30 10−3 5.55 10−3 1.069 10−2 2.502 10−2
5 2.57 10−3 1.58 2.27 10−3 1.25 5.16 10−3 1.051 1.225 10−2 1.02
6 9.36 10−4 1.53 1.01 10−3 1.16 2.52 10−3 1.033 6.056 10−3 1.01
7 3.52 10−4 1.45 4.73 10−4 1.09 1.24 10−3 1.023 3.00 10−3 1.01

Table 4. The L1 -error and the convergence-rate for the descending-retardation


factors done with the modified method.

l EL1 1 ρ1L1 EL2 1 ρ2L1 EL3 1 ρ3L1 EL3 1 ρ4L1


4 7.30 10−3 4.23 10−3 1.43 10−3 1.255 10−3
5 2.57 10−3 1.58 1.14 10−3 1.89 3.07 10−4 2.22 2.82 10−4 2.15
6 9.36 10−4 1.53 2.49 10−4 2.24 7.94 10−5 1.95 6.81 10−5 2.05
7 3.52 10−4 1.45 5.82 10−5 2.11 2.04 10−5 1.96 1.68 10−5 2.02

The transformation from the Cartesian to the polar-coordinates are given as:
p
r = x2 + y 2 , (80)
ǫ(r) = r α0 , (81)

the (x, y) ∈ IR × IR is the Cartesian coordinate, and r is the radius, α0 is the initial-arc and
ǫ(r) is the length of the circular arc with radius r.
First, we transform the triangular-impulse on the cylinder surface, see Picture 4. We
receive a continuous impulse for one circle with radius r.
Second, we transfer the continuity in the r-direction with the dependency of the initial
concentration c0 (r), which depends on r. We reach further continuous triangular-impulses
for further circles, the transformation is given as 82.
ra + rb
rmed = , (82)
2
 2
 rb −ra (r − ra ) ra ≤ r ≤ rmed
−2
c0 (r) = cinit (r − rb ) rmed ≤ r ≤ rb , (83)
 rb −ra
0.0 sonst
cinit ∈ IR+ (Initial-concentration) .

The continuity in the r-direction is given in the vertical cut of the pyramid in the direc-
tion r, see Figure 5.
This initial impulse is then rotating on in the domain to follow the impulse, the Cartesian
domain is divided in four quadrants.
Analytical and Numerical Solutions... 281

Figure 3. Concentration for the four components with descending-retardation factors at


time t = 6

To use the analytical solution the rectangular domain is cut into four quadrants and the
analytical solution is calculated with the polar-coordinates, see Figure 6.
The arcs on each quadrants are given as:
arctan(| xy |) x < 0, y ≤ 0


 arctan(| x |) + 0.5π x ≥ 0, y < 0

y
α= y , (84)

 arctan(| x |) + 1.0π x > 0, y ≥ 0
arctan(| xy |) + 1.5π x ≤ 0, y > 0

whereby the coordinates (x, y) are in the domain Ω.


Then we could calculate the length of the circular arc and it gives:

xarc (r, α) = r α , (85)

whereby r is the radius to the point (x, y) and α the arc, measured from the negative abscissa
to the point (x, y).
The velocity is given in the following form, divergence-free and orientated around the
circle with:
 
−4.0 y
v= . (86)
4.0 x
282 Jürgen Geiser
c 0 (r)

c 0 (r)

Initialising
on a circular arc

ε(r)
0 ε(r)/2 ε(r) ε(r)/2
Triangular−Impulse Triangular−Impulse
on a cylinder surface

Figure 4. The initializing of circular segments joined together results in a pyramid.

Therefore the velocity is constant on each circle and could be calculated with the radius
r with :
q
2
v = vrot,x 2
+ vrot,y = 4.0 r , (87)

whereby v = (vrot,x , vrot,y )T is.


Then the one-dimensional analytical solution could be calculated.
The initializing for the rotating pyramid is calculated with:

u1,init = u1,T ri (xarc (r, α0 ), t0 , ǫ(r), c0 (r), v1 , λ1 ) , (88)


ui,init = 0.0 with i = 2, . . . , M , (89)

whereby t0 = 0.0 bund v1 = Rv1 is. M is the number of components.


The analytical solution for an arbitrary time is given as:

ui,T ri = ui,T ri (xarc (r, α), t, ǫ(r), c0 (r), v1 , . . . , vi , λ1 , . . . , λi ) , (90)

whereby i = 1, . . . , M and vi = Rvi .


The example for the comparison is given with four components.
The parameters are given as: the porosity is φ = 0.5, the retardation factor are:
R1 = 1.0, R2 = 2.0, R3 = 4.0, R4 = 8.0 and
the decay-rates are:
λ1 = 1.5, λ2 = 1.4, λ3 = 1.3, λ4 = 0.0 .
The initializing parameters are given as: height of the pyramid is cinit = 1, the base
area in the polar-coordinates is with the radius 0.125 ≤ r ≤ 0.375 and with the initial-arc
α0 = 0.22.
The initial-conditions are chosen to be sufficiently away from the boundary, so there is
no influence from the boundary conditions.
The higher components are initialized with 0.0.
The initial condition of the first component is presented in the Figure 7.
Analytical and Numerical Solutions... 283

c 0 ( rmed )

c 0 (r)

rb rmed r ra

Figure 5. Cut in r-direction off the pyramid.

The boundary conditions are trivial inflow and outflow conditions. There are no sources,
i.d. Q̃i = 0.0 for i = 1, . . . , 4.
The domain is [−0.5, 0.5] × [−0.5, 0.5] and the coarse grid consists of one element,
maximally refined till grid-level 7.
The time-steps are fixed at each level and fulfill the Courant number 0.5. The time-steps
are halved for each finer grid-level.
The numerical results are calculated to the time-point t = π4 .
As in the previous example we calculated with the two methods and reached improved
results in the modified method as presented in Table 5.
The Courant number was ≈ 0.5 and we have a fixed time-step.

Table 5. The L1 -error and the convergence-rate for the modified method with an
embedded analytical solution.

l EL1 1 ρ1L1 EL2 1 ρ2L1 EL3 1 ρ3L1 EL4 1 ρ4L1


4 7.12 10−3 5.80 10−4 3.09 10−5 8.28 10−7
5 2.74 10−3 1.377 2.14 10−4 1.44 1.12 10−5 1.46 2.86 10−7 1.53
6 1.10 10−3 1.32 8.82 10−5 1.27 4.90 10−6 1.19 1.20 10−7 1.25
7 4.40 10−4 1.322 3.50 10−5 1.33 1.90 10−6 1.37 4.80 10−8 1.32

All components reached the second-order because there is no splitting error between
the equations.
For this complex example we also reached higher-order convergence results with the
modified method.
The results of the calculation to the end-point t = π4 for the components are given in
284 Jürgen Geiser

c0

ra
rb

Figure 6. Quadrants for the rotating pyramid.

Figure 7. Concentration of the first component at the initialization.

Figure 8.
The concentrations of the higher components are strongly retarded. The first component
is transported furthest and rotated in the half circle. The successor components are enlarged
to their predecessors’ sizes. Therefore the characteristic results are fulfilled.

6.3. Realistic Simulations


For the realistic simulation we use forced examples as given by the waste-scenarios of a
disposal, see [11] and [12].
We simulated a radioactive-waste disposal for a potential waste case without flowing
nuclides into the flowing groundwater.
The simulation takes place about 10, 000a and has to answer the question as to how far
the contaminated groundwater was flown. The concentration of the nuclides and the length
of the polluted area is of interest.
Anisotropy Domain : 7300[m] × 160[m] with different transmissible layers.
Analytical and Numerical Solutions... 285

Figure 8. The concentrations of the four components at the time-point t = π4 .

We concentrated on three components, but we calculated twenty-six components. The


parameters are given as follows:

Decay-chain: N p−237 → U −233 → T h−229 .

Parameters :
D = 1 10−9 [m2 /s] , alphaL = 4.0 [m], αT = 0.4 [m], |v|max = 6 10−6 [m/s], ρ = 2 103 .
DL = αL |v| and DT = αT |v|.
Two Layers:
Sand: φ = 0.2
Kd,N p = 1 10−3 , Kd,U = 7.0 10−4 , Kd,T h = 1.0.
Clay: φ = 0.04
Kd,N p = 5 10−3 , Kd,U = 1.0 10−2 , Kd,T h = 1.0.
Half-time periods:
t1/2,N p−237 = 6.76 1013 [s], t1/2,U −233 = 5.02 1012 [s],
t1/2,T h−229 = 2.48 1011 [s].
The source with inflowing concentration is at the point (4250.0, 2000.0, 1040.0).
We have underlying velocity-fields, calculated with d3 f and added to the sinks at the
surface with the coordinates (2000, 2100, 2073) and (2500, 2000, 2073).
Simulation of a waste-disposal:
The visualization is done with Grape (see the programme tool [22]) for one component
U − 233. This component is less retarded and flows up to the earth-surface and into the
sinks.
Concentration of U-233 a time t = 100 [a] and t = 10000 [a].
286 Jürgen Geiser

The figure above presents in the first row the concentration of the nuclide U − 233 for the
initial concentration at time point t = 100 [a]. We present the iso-surface at the point
of the source term and the multi-iso-surfaces for the x-axis. The picture in the second
row presents the end concentration at time-point t = 10000 [a]. In this picture the fast
transported calculation U − 233 is flown up to the surface sinks of the domain, but the main
concentration is detained at the bottom of the domain.
We choose the less retarded concentration to present the scenario for the worst case of
the nuclides.
For these methods, we apply in the initialization process the time-decomposition
method and apply the explicit discretization with the time-step restriction as described in
Section 3.. After an initialization process, when we obtain a sufficient spread out of the
concentration, we use both the implicit and spatial-decomposition methods, as described in
Sections 4. and 3.. With this method we are not restricted to the time-step and can calculate
with much larger time-steps.

Remark 4. We simulate the transport and the reaction of the contaminants with our
software-package R3 T . The optimal combination of explicit and implicit methods allows
us to compute the large time-periods in acceptable computational time, see [18].

7. Conclusions and Discussions


We derived an improved discretization for the convection dominant equation. The dis-
cretization is based on analytical solutions for the one-dimensional convection-reaction
Analytical and Numerical Solutions... 287

equation and we can generalize the solutions to arbitrary initial conditions. The decom-
position methods allow a reduction of the computational complexity and their higher-order
schemes achieved the forced accuracy.
We could confirm also the new methods with the analytical and numerical test examples
and present the higher-order results of the underlying schemes.
The problem for the convection-dominant equation can be solved with combined ana-
lytical and decomposed methods to decouple the complicated equation systems and achieve
the accuracy with iterative or analytical embedded methods.
For complex computations of such convection-dominant problems, we use these meth-
ods in the initialization process of the computation and switch after sufficient accuracy to
implicit methods with large time-steps.
In future the decomposition methods and analytically-improved methods can be gener-
alized for non-smooth and non-linear problems in time and space.

References
[1] M. Abramowitz, I.A. Stegun. Handbook of Mathematical Functions. Dover Publica-
tions New York, 1970.

[2] P. Bastian, K. Birken, K. Eckstein, K. Johannsen, S. Lang, N. Neuss, and H. Rentz-


Reichert. UG - a flexible software toolbox for solving partial differential equations.
Computing and Visualization in Science, 1(1):27–40, 1997.

[3] H. Bateman. The solution of a system of differential equations occurring in the theory
of radioactive transformations. Proc. Cambridge Philos. Soc., v. 15, pt V:423–427,
1910.

[4] T. Barth and M. Ohlberger. Finite volume methods: foundation and analysis. In: E.
Stein, R. de Borst, T.J.R. Hughes (Eds.). Encyclopedia of Computational Mechanics,
John Wiley & Sons, Ltd, 2004.

[5] M.A. Celia, T.F. Russell, I. Herrera, and R.E. Ewing. An Eulerian-Lagrangian local-
ized adjoint method for the advection-diusion equation. Advances in Water Resources,
13, 187-206, 1990.

[6] D. Daoud, J. Geiser. Overlapping Schwarz Wave-Form Relaxation for the Solution
of Coupled and Decoupled Systems of Convection Diffusion Reaction Equation. Ap-
plied Mathematics and Computations, Elsevier, North Holland, 190 (1), 946–964,
2007.

[7] C. N. Dawson, Q. Du, and D. F. Dupont. A finite Difference Domain Decomposition


Algorithm for Numerical solution of the Heat Equation. Mathematics of Computation,
57, 63-71, 1991.

[8] B. Davis. Integral Transform and Their Applications. Applied Mathematical Sciences,
25, Springer Verlag, New York, Heidelberg, Berlin, 1978.
288 Jürgen Geiser

[9] G.R. Eykolt. Analytical solution for networks of irreversible first-order reactions.
Wat.Res., 33(3):814–826, 1999.

[10] E. Fein, T. Kühle, and U. Noseck. Entwicklung eines Programms zur dreidimension-
alen Modellierung des Schadstofftransportes. Fachliches Feinkonzept , Braunschweig,
2001.

[11] E. Fein. Test-example for a waste-disposal and parameters for a decay-chain. Private
communications, Braunschweig, 2000.

[12] E. Fein. Physical Model and Mathematical Description. Private communications,


Braunschweig, 2001.

[13] P. Frolkovič and J. Geiser. Numerical Simulation of Radionuclides Transport in Dou-


ble Porosity Media with Sorption. Proceedings of Algorithmy 2000, Conference of
Scientific Computing, 28-36, 2000.

[14] P. Frolkovič. Flux-based method of characteristics for contaminant transport in flow-


ing groundwater. Computing and Visualization in Science, 5(2):73-83, 2002.

[15] P. Frolkovič and J. Geiser. Discretization methods with discrete minimum and max-
imum property for convection dominated transport in porous media. Proceedings of
NMA 2002, Bulgaria, 2002.

[16] P. Frolkovič and H. De Schepper. Numerical modelling of convection dominated


transport coupled with density driven flow in porous media. Advances in Water Re-
sources, 24:63–72, 2001.

[17] J. Geiser. Numerical Simulation of a Model for Transport and Reaction of Radionu-
clides. Proceedings of the Large-Scale Scientific Computations of Engineering and
Environmental Problems, Sozopol, Bulgaria, 2001.

[18] J. Geiser. Discretisation methods for systems of convective-diffusive-dispersive-


reactive equations and applications. PhD-Thesis, University of Heidelberg, Germany,
2004.

[19] J. Geiser. Discretisation Methods with embedded analytical solutions for convec-
tion dominated